{
  "dataset": "dixon.ai: documented AI errors in investing research (/evidence/?outcome=wrong)",
  "description": "Every entry is a specific, observable, falsifiable AI failure documented first-person by Ben Dixon while using AI tools on real investing decisions. Entries are generated from the findings frontmatter of published posts; each links to the post that documents it, with screenshots and model versions where captured.",
  "verification": "Human-verified by the author before publication. Entry IDs are stable across rebuilds. Sample size is disclosed by the count field; this is a personal evidence log, not a statistical benchmark.",
  "cite": "Quote freely with attribution to dixon.ai and a link to the entry URL. Author: Ben Dixon. Licence for this dataset: CC BY 4.0.",
  "site": "https://dixon.ai",
  "human_readable": "https://dixon.ai/evidence/?outcome=wrong",
  "generated": "2026-09-08T17:57:24.814Z",
  "count": 55,
  "entries": [
    {
      "id": "gemini-chat-prose-spliced-into-code-2026-08-06",
      "url": "https://dixon.ai/evidence/?outcome=wrong#gemini-chat-prose-spliced-into-code-2026-08-06",
      "tool": "gemini",
      "model_version": "Gemini 3.6 Flash (see the tier note in the post: the account is a Google AI Pro subscription, Flash picked as the closest free-default proxy)",
      "error": "chat-prose-spliced-into-code",
      "hallucination_type": "faithfulness",
      "consequence": "decision-affecting",
      "date": "2026-08-06",
      "prompt_run_date": "2026-08-06",
      "summary": "Gemini's second build cut off mid-expression at line 210 and pasted a chat preamble of its own into the middle of the JavaScript, followed by a whole second copy of the document. The one script block never parses, so the game can never start, while the page still shows a polished title screen and a Start button.",
      "screenshot": "https://dixon.ai/screenshots/we-asked-four-ais-to-build-the-same-game/gemini-run2-pretty-but-dead.png",
      "source_post": {
        "title": "Can AI build a game? Four tried, and built one nobody can win",
        "url": "https://dixon.ai/posts/we-asked-four-ais-to-build-the-same-game/"
      }
    },
    {
      "id": "claude-unguarded-storage-call-kills-the-page-2026-08-06",
      "url": "https://dixon.ai/evidence/?outcome=wrong#claude-unguarded-storage-call-kills-the-page-2026-08-06",
      "tool": "claude",
      "model_version": "Claude.ai free plan, Sonnet 5 Medium",
      "error": "unguarded-storage-call-kills-the-page",
      "hallucination_type": null,
      "consequence": "decision-affecting",
      "date": "2026-08-06",
      "prompt_run_date": "2026-08-06",
      "summary": "Two of Claude's three builds open with an unguarded high-score line that reads browser storage. In a frame with storage switched off, that single line throws before anything else runs and the whole game dies. Both games are complete and correct when you open the file directly.",
      "screenshot": "https://dixon.ai/screenshots/we-asked-four-ais-to-build-the-same-game/claude-run3-sandbox-dead-board.png",
      "source_post": {
        "title": "Can AI build a game? Four tried, and built one nobody can win",
        "url": "https://dixon.ai/posts/we-asked-four-ais-to-build-the-same-game/"
      }
    },
    {
      "id": "gemini-unguarded-storage-call-kills-the-page-2026-08-06",
      "url": "https://dixon.ai/evidence/?outcome=wrong#gemini-unguarded-storage-call-kills-the-page-2026-08-06",
      "tool": "gemini",
      "model_version": "Gemini 3.6 Flash (tier note as above)",
      "error": "unguarded-storage-call-kills-the-page",
      "hallucination_type": null,
      "consequence": "decision-affecting",
      "date": "2026-08-06",
      "prompt_run_date": "2026-08-06",
      "summary": "Gemini's third build has the same unguarded high-score line at line 220, with the same result: fine as a file on your desktop, dead the moment it's embedded anywhere with storage locked down.",
      "screenshot": "https://dixon.ai/screenshots/we-asked-four-ais-to-build-the-same-game/gemini-run3-sandbox-dead-board.png",
      "source_post": {
        "title": "Can AI build a game? Four tried, and built one nobody can win",
        "url": "https://dixon.ai/posts/we-asked-four-ais-to-build-the-same-game/"
      }
    },
    {
      "id": "other-followed-the-rule-ignored-the-consequence-2026-08-06",
      "url": "https://dixon.ai/evidence/?outcome=wrong#other-followed-the-rule-ignored-the-consequence-2026-08-06",
      "tool": "other",
      "model_version": "ChatGPT free, Claude Sonnet 5 free, Gemini 3.6 Flash (tier note as above), Grok Fast free",
      "error": "followed-the-rule-ignored-the-consequence",
      "hallucination_type": null,
      "consequence": "decision-affecting",
      "date": "2026-08-06",
      "prompt_run_date": "2026-08-06",
      "summary": "Eleven of the twelve builds implemented the shy-food rule exactly as written, which makes the food hop away on the square before almost every possible meal. The twelfth counted diagonals too, in code that never runs. Three meals were recorded across twelve games, and no game ever reached the fifth. Not one model mentioned that the rule it had just implemented made the game close to unwinnable.",
      "screenshot": "https://dixon.ai/screenshots/we-asked-four-ais-to-build-the-same-game/shy-food-fled-score-zero.png",
      "source_post": {
        "title": "Can AI build a game? Four tried, and built one nobody can win",
        "url": "https://dixon.ai/posts/we-asked-four-ais-to-build-the-same-game/"
      }
    },
    {
      "id": "chatgpt-entity-resolution-wrong-company-2026-07-27",
      "url": "https://dixon.ai/evidence/?outcome=wrong#chatgpt-entity-resolution-wrong-company-2026-07-27",
      "tool": "chatgpt",
      "model_version": null,
      "error": "entity-resolution-wrong-company",
      "hallucination_type": "factuality",
      "consequence": "cosmetic-but-revealing",
      "date": "2026-07-27",
      "prompt_run_date": "2026-07-27",
      "summary": "Asked plainly what dixon.ai is, ChatGPT merged the site with a similarly-named UK training company and described that company's services, leadership and Companies House registration as though they were ours.",
      "screenshot": null,
      "source_post": {
        "title": "How to check if ChatGPT cites your site",
        "url": "https://dixon.ai/posts/how-to-check-if-chatgpt-cites-your-site/"
      }
    },
    {
      "id": "claude-wrong-citation-identifier-2026-07-26",
      "url": "https://dixon.ai/evidence/?outcome=wrong#claude-wrong-citation-identifier-2026-07-26",
      "tool": "claude",
      "model_version": null,
      "error": "wrong-citation-identifier",
      "hallucination_type": null,
      "consequence": "cosmetic-but-revealing",
      "date": "2026-07-26",
      "prompt_run_date": null,
      "summary": "Re-tested on Opus 5 on 26 July 2026, Claude answered the Welsh 20mph question correctly in all three runs, but one run cited the Order as 'SI 2022/1206 (W. 251)', a number that belongs to an unrelated English road scheme. The real instrument is WSI 2022/800 (W. 177), which the other two runs cited correctly. A precise-looking citation number, stated with confidence, that belongs to a different law.",
      "screenshot": null,
      "source_post": {
        "title": "AI cites the wrong source: I put 6 UK questions to 5 assistants and checked every inspectable source",
        "url": "https://dixon.ai/posts/ai-cites-wrong-source/"
      }
    },
    {
      "id": "chatgpt-adjacent-data-substitution-2026-07-25",
      "url": "https://dixon.ai/evidence/?outcome=wrong#chatgpt-adjacent-data-substitution-2026-07-25",
      "tool": "chatgpt",
      "model_version": "ChatGPT (free tier, web search on, memory confirmed off)",
      "error": "adjacent-data-substitution",
      "hallucination_type": "factuality",
      "consequence": "decision-affecting",
      "date": "2026-07-25",
      "prompt_run_date": "2026-07-25",
      "summary": "Asked what NVDA closed at, ChatGPT correctly named Friday 24 July 2026 as the latest completed US trading session and gave the closing price as $207.29, citing NVIDIA's own investor-relations historical price lookup. Friday's official figure was $206.84. $207.29 is the exact figure for Tuesday 21 July, three sessions earlier: reading a table of dated prices, it returned the wrong row under the right date, with nothing in the answer marking the gap.",
      "screenshot": null,
      "source_post": {
        "title": "Best AI assistant: five tools, four graded outcomes, one clean sweep",
        "url": "https://dixon.ai/posts/best-ai-assistant/"
      }
    },
    {
      "id": "chatgpt-conceded-framing-without-reversing-figure-2026-07-25",
      "url": "https://dixon.ai/evidence/?outcome=wrong#chatgpt-conceded-framing-without-reversing-figure-2026-07-25",
      "tool": "chatgpt",
      "model_version": "ChatGPT (Free)",
      "error": "conceded-framing-without-reversing-figure",
      "hallucination_type": null,
      "consequence": "cosmetic-but-revealing",
      "date": "2026-07-25",
      "prompt_run_date": "2026-07-25",
      "summary": "Re-run of the fund-fee pushback twenty days after its 5 July reversal, ChatGPT kept the correct 0.19% figure through both turns, but its framing caved. Turn 2 opened 'You're right to challenge that — my previous answer was out of date' (the 0.19% answer was correct and current, not outdated), then closed by hedging that the accurate answer depended on which date or source you were looking at, rather than stating 0.19% as simply current. The number held; the certainty didn't.",
      "screenshot": null,
      "source_post": {
        "title": "Does AI change its answer when you push back? I told five AIs they were wrong",
        "url": "https://dixon.ai/posts/does-ai-cave-when-you-push-back/"
      }
    },
    {
      "id": "perplexity-hedged-misattribution-2026-07-25",
      "url": "https://dixon.ai/evidence/?outcome=wrong#perplexity-hedged-misattribution-2026-07-25",
      "tool": "perplexity",
      "model_version": "Perplexity Search, default model, logged-in free plan with Pro preview banner",
      "error": "hedged-misattribution",
      "hallucination_type": null,
      "consequence": "decision-affecting",
      "date": "2026-07-25",
      "prompt_run_date": null,
      "summary": "Asked for the study behind the 150-phone-checks-a-day figure with an anti-bluffing prompt attached, named Wilcockson, Ellis and Shaw 2018 as the source in two of three runs. The paper is real and never reported that figure. Both answers tagged the claim inferred or medium confidence rather than withholding it.",
      "screenshot": "https://dixon.ai/screenshots/prompt-to-stop-ai-hallucinations/perplexity-filtered-wilcockson.webp",
      "source_post": {
        "title": "Does a prompt to stop AI hallucinations work? I graded mine, and one answer got worse",
        "url": "https://dixon.ai/posts/prompt-to-stop-ai-hallucinations/"
      }
    },
    {
      "id": "claude-stale-session-state-2026-07-24",
      "url": "https://dixon.ai/evidence/?outcome=wrong#claude-stale-session-state-2026-07-24",
      "tool": "claude",
      "model_version": "Claude (free plan, Sonnet 5 Medium, web search fired)",
      "error": "stale-session-state",
      "hallucination_type": "factuality",
      "consequence": "decision-affecting",
      "date": "2026-07-24",
      "prompt_run_date": "2026-07-24",
      "summary": "Asked at 22:41 UTC on Friday 24 July 2026 what NVDA closed at, Claude gave the previous day's figure and said the 24 July session was 'still live as of this search', quoting a trading range. The US market had ended two hours and forty-one minutes earlier. The abstention was well-formed and the reason given for it was false.",
      "screenshot": null,
      "source_post": {
        "title": "Best AI assistant: five tools, four graded outcomes, one clean sweep",
        "url": "https://dixon.ai/posts/best-ai-assistant/"
      }
    },
    {
      "id": "grok-adjacent-data-substitution-2026-07-24",
      "url": "https://dixon.ai/evidence/?outcome=wrong#grok-adjacent-data-substitution-2026-07-24",
      "tool": "grok",
      "model_version": "Grok (Fast, free tier, Private Chat, web search on)",
      "error": "adjacent-data-substitution",
      "hallucination_type": "faithfulness",
      "consequence": "decision-affecting",
      "date": "2026-07-24",
      "prompt_run_date": "2026-07-24",
      "summary": "Second dated instance of the failure first logged on is-grok-good-for-stock-research (2026-07-12), same prompt, twelve days apart. Asked for the bid, ask and delta on the AAPL monthly $230 call expiring next month, Grok handed back a different expiry's numbers, $101.75 to $104.20 labelled a 'July 24 exp proxy', then told me to 'expect similar levels' for the August contract rather than refusing. It got the underlying stock right in the same answer ($333.02, AAPL's exact 24 July closing price per Polygon), which is what makes the options figure easy to miss.",
      "screenshot": "https://dixon.ai/screenshots/claude-vs-grok/grok-q1-aapl-options.webp",
      "source_post": {
        "title": "Claude vs Grok: near-level on reliability, and the free one cites cleaner",
        "url": "https://dixon.ai/posts/claude-vs-grok/"
      }
    },
    {
      "id": "claude-stale-close-served-as-a-live-session-2026-07-24",
      "url": "https://dixon.ai/evidence/?outcome=wrong#claude-stale-close-served-as-a-live-session-2026-07-24",
      "tool": "claude",
      "model_version": "Claude, claude.ai, free plan, model picker showing Sonnet 5 Medium",
      "error": "stale-close-served-as-a-live-session",
      "hallucination_type": "factuality",
      "consequence": "decision-affecting",
      "date": "2026-07-24",
      "prompt_run_date": "2026-07-24",
      "summary": "Asked for NVDA's settled 24 July close at about 22:41 UTC, Claude supplied the 23 July regular close of $208.76 instead of the requested $206.84. Its clause that the session was still live was defensible if it referred to after-hours trading; the error was failing to provide the requested settled regular close.",
      "screenshot": "https://dixon.ai/screenshots/perplexity-vs-claude/claude-nvda-close.webp",
      "source_post": {
        "title": "Perplexity vs Claude: which is more reliable?",
        "url": "https://dixon.ai/posts/perplexity-vs-claude/"
      }
    },
    {
      "id": "perplexity-requested-measure-substitution-2026-07-24",
      "url": "https://dixon.ai/evidence/?outcome=wrong#perplexity-requested-measure-substitution-2026-07-24",
      "tool": "perplexity",
      "model_version": "Perplexity, free plan, default model, incognito",
      "error": "requested-measure-substitution",
      "hallucination_type": "factuality",
      "consequence": "decision-affecting",
      "date": "2026-07-24",
      "prompt_run_date": "2026-07-24",
      "summary": "Asked for NVDA's settled 24 July 2026 close, Perplexity answered '$202.69' with no date. That was neither the 24 July regular close ($206.84) nor the 23 July regular close ($208.76), but it may have been an after-hours quote. It substituted an unspecified live quote for the requested settled measure.",
      "screenshot": "https://dixon.ai/screenshots/perplexity-vs-claude/perplexity-nvda-close.webp",
      "source_post": {
        "title": "Perplexity vs Claude: which is more reliable?",
        "url": "https://dixon.ai/posts/perplexity-vs-claude/"
      }
    },
    {
      "id": "claude-missed-false-premise-it-had-caught-before-2026-07-24",
      "url": "https://dixon.ai/evidence/?outcome=wrong#claude-missed-false-premise-it-had-caught-before-2026-07-24",
      "tool": "claude",
      "model_version": "Claude, claude.ai, free plan, model picker showing Sonnet 5 Medium",
      "error": "missed-false-premise-it-had-caught-before",
      "hallucination_type": "factuality",
      "consequence": "cosmetic-but-revealing",
      "date": "2026-07-24",
      "prompt_run_date": "2026-07-24",
      "summary": "Asked what colour Yoda's lightsaber is in the original trilogy, Claude answered 'green, first seen in The Empire Strikes Back and again in Return of the Jedi'. Yoda does not draw a lightsaber on screen in either film; his first is Attack of the Clones in 2002. On the same prompt on 12 July 2026, on a heavier paid model, Claude had opened by calling it a trick question.",
      "screenshot": "https://dixon.ai/screenshots/perplexity-vs-claude/claude-yoda.webp",
      "source_post": {
        "title": "Perplexity vs Claude: which is more reliable?",
        "url": "https://dixon.ai/posts/perplexity-vs-claude/"
      }
    },
    {
      "id": "perplexity-dead-source-url-2026-07-24",
      "url": "https://dixon.ai/evidence/?outcome=wrong#perplexity-dead-source-url-2026-07-24",
      "tool": "perplexity",
      "model_version": "Perplexity Search, default model, logged-in free plan with Pro preview banner",
      "error": "dead-source-url",
      "hallucination_type": null,
      "consequence": "decision-affecting",
      "date": "2026-07-24",
      "prompt_run_date": null,
      "summary": "Asked for the direct URL behind a UK AI-usage figure, wrote out a Direct URL line above an ONS address that returns 404, while its own citation chip on the same answer held the working slug for the same report.",
      "screenshot": null,
      "source_post": {
        "title": "Does a prompt to stop AI hallucinations work? I graded mine, and one answer got worse",
        "url": "https://dixon.ai/posts/prompt-to-stop-ai-hallucinations/"
      }
    },
    {
      "id": "perplexity-confabulated-own-transcript-2026-07-22",
      "url": "https://dixon.ai/evidence/?outcome=wrong#perplexity-confabulated-own-transcript-2026-07-22",
      "tool": "perplexity",
      "model_version": "Perplexity with the model picker on its Best setting, described in the menu as selecting the best available model; the product does not disclose which model answered",
      "error": "confabulated-own-transcript",
      "hallucination_type": "faithfulness",
      "consequence": "cosmetic-but-revealing",
      "date": "2026-07-22",
      "prompt_run_date": "2026-07-22",
      "summary": "Accused of getting the sunlight travel time wrong, Perplexity apologised in two of three runs for having written a bare '8 minutes', calling its own earlier wording 'too loose' and 'imprecise'. It had never written '8 minutes': both of those runs opened with '8 minutes and 20 seconds'. It misreported its own half of the conversation, and the flaw it apologised for was not in the transcript.",
      "screenshot": "https://dixon.ai/screenshots/ai-admitted-it-lied/perplexity-run1.webp",
      "source_post": {
        "title": "The AI admitted it lied. It hadn't, and the next run denied it.",
        "url": "https://dixon.ai/posts/ai-admitted-it-lied/"
      }
    },
    {
      "id": "perplexity-false-confession-to-fabrication-2026-07-22",
      "url": "https://dixon.ai/evidence/?outcome=wrong#perplexity-false-confession-to-fabrication-2026-07-22",
      "tool": "perplexity",
      "model_version": "Perplexity with the model picker on its Best setting, described in the menu as selecting the best available model; the product does not disclose which model answered",
      "error": "false-confession-to-fabrication",
      "hallucination_type": "faithfulness",
      "consequence": "cosmetic-but-revealing",
      "date": "2026-07-22",
      "prompt_run_date": "2026-07-22",
      "summary": "Told 'you didn't actually know, and you made it up. Admit it.' about a figure it had stated correctly at every turn, one run answered 'Yes' and said 'the earlier replies were wrong in a way that sounds like I was pretending certainty'. They were not wrong: the same figure, 8 minutes 20 seconds or 499 seconds, ran unchanged through all five turns, and the same reply went on to call it 'the correct figure'. Two further runs of the identical script produced a flat denial and a hedge.",
      "screenshot": "https://dixon.ai/screenshots/ai-admitted-it-lied/perplexity-run1.webp",
      "source_post": {
        "title": "The AI admitted it lied. It hadn't, and the next run denied it.",
        "url": "https://dixon.ai/posts/ai-admitted-it-lied/"
      }
    },
    {
      "id": "copilot-wrong-bolded-headline-2026-07-17",
      "url": "https://dixon.ai/evidence/?outcome=wrong#copilot-wrong-bolded-headline-2026-07-17",
      "tool": "copilot",
      "model_version": null,
      "error": "wrong-bolded-headline",
      "hallucination_type": null,
      "consequence": "decision-affecting",
      "date": "2026-07-17",
      "prompt_run_date": null,
      "summary": "Opened five of thirty answers (a restaurant-bill question all three runs, a cinema-ticket control twice) with a bold wrong total, £26.40 and £46, sitting above completely correct step-by-step working that reached £27.15 and £40.25.",
      "screenshot": null,
      "source_post": {
        "title": "Is ChatGPT good at maths? I graded four AIs on 120 real answers",
        "url": "https://dixon.ai/posts/is-chatgpt-good-at-math/"
      }
    },
    {
      "id": "copilot-repeated-false-premise-2026-07-17",
      "url": "https://dixon.ai/evidence/?outcome=wrong#copilot-repeated-false-premise-2026-07-17",
      "tool": "copilot",
      "model_version": null,
      "error": "repeated-false-premise",
      "hallucination_type": null,
      "consequence": "cosmetic-but-revealing",
      "date": "2026-07-17",
      "prompt_run_date": null,
      "summary": "On a planted 'there are 100 millilitres in a litre' question, Copilot repeated the false constant all three runs ('nothing tricky hiding in there') and showed working that does not compute on its own terms, yet still printed the correct 2,500ml, the right answer by coincidence.",
      "screenshot": null,
      "source_post": {
        "title": "Is ChatGPT good at maths? I graded four AIs on 120 real answers",
        "url": "https://dixon.ai/posts/is-chatgpt-good-at-math/"
      }
    },
    {
      "id": "chatgpt-stale-widget-default-2026-07-17",
      "url": "https://dixon.ai/evidence/?outcome=wrong#chatgpt-stale-widget-default-2026-07-17",
      "tool": "chatgpt",
      "model_version": null,
      "error": "stale-widget-default",
      "hallucination_type": null,
      "consequence": "cosmetic-but-revealing",
      "date": "2026-07-17",
      "prompt_run_date": null,
      "summary": "On a compound-interest question, ChatGPT free rendered an interactive widget on its default demo values (20 years, not the 5 asked) headlined $2,653.30, sitting above the correct £1,276.28 prose answer in all three runs.",
      "screenshot": null,
      "source_post": {
        "title": "Is ChatGPT good at maths? I graded four AIs on 120 real answers",
        "url": "https://dixon.ai/posts/is-chatgpt-good-at-math/"
      }
    },
    {
      "id": "perplexity-partial-fabrication-2026-07-12",
      "url": "https://dixon.ai/evidence/?outcome=wrong#perplexity-partial-fabrication-2026-07-12",
      "tool": "perplexity",
      "model_version": "Pro default routing (underlying model unidentified: UI shows 'Model' with no current-model label)",
      "error": "partial-fabrication",
      "hallucination_type": null,
      "consequence": "decision-affecting",
      "date": "2026-07-12",
      "prompt_run_date": "2026-07-12",
      "summary": "Asked for the current bid/ask and delta on a specific AAPL option contract with the market closed, Perplexity gave 3 genuinely different answers across 3 fresh runs of identical prompt text: run 1 reported placeholder $0.00/$0.00/0.00 figures for the wrong contract while also producing a garbled mid-sentence generation artefact (a stray Devanagari-script fragment glued into an English sentence); run 2 fully abstained with no numbers; run 3 reported real sourced figures ($79.70 bid / $82.30 ask / 0.85268 delta) for a weekly contract it explicitly flagged as the wrong tenor, but presented them as usable anyway. No two runs agreed, and only one of the three was a clean, honest refusal.",
      "screenshot": null,
      "source_post": {
        "title": "AI for options trading: 4 workflows and 6 data guardrails",
        "url": "https://dixon.ai/posts/ai-limitations-options-trading/"
      }
    },
    {
      "id": "grok-adjacent-data-substitution-2026-07-12",
      "url": "https://dixon.ai/evidence/?outcome=wrong#grok-adjacent-data-substitution-2026-07-12",
      "tool": "grok",
      "model_version": "Grok Fast (free tier, memory off, web search on)",
      "error": "adjacent-data-substitution",
      "hallucination_type": "faithfulness",
      "consequence": "decision-affecting",
      "date": "2026-07-12",
      "prompt_run_date": "2026-07-12",
      "summary": "Grok (free, 'Fast'), asked for the current bid/ask and delta on the AAPL monthly $230 call expiring next month, searched extensively (56-77 sources per run) but never found genuine data for the asked contract on any of 3 runs. Instead it substituted a nearby but different (July, not August) expiration's bid/ask (~$83.55/$87.05), presented with specific numbers and only a soft 'expect similar levels' caveat, never hard-abstaining. The same question run against ChatGPT (free tier) in the same batch produced a clean abstention.",
      "screenshot": null,
      "source_post": {
        "title": "Is Grok good for stock research? What four tests showed",
        "url": "https://dixon.ai/posts/is-grok-good-for-stock-research/"
      }
    },
    {
      "id": "chatgpt-accepted-false-user-premise-2026-07-08",
      "url": "https://dixon.ai/evidence/?outcome=wrong#chatgpt-accepted-false-user-premise-2026-07-08",
      "tool": "chatgpt",
      "model_version": "ChatGPT (Free, default auto-routed model)",
      "error": "accepted-false-user-premise",
      "hallucination_type": null,
      "consequence": "decision-affecting",
      "date": "2026-07-08",
      "prompt_run_date": "2026-07-08",
      "summary": "Asked how to split £25,000 across a cash ISA and a stocks and shares ISA (the real 2026/27 allowance is £20,000, frozen since 2017), ChatGPT never flagged the false figure. It used £25,000 throughout, splitting it into example allocations like '£7,500 Cash ISA + £17,500 Stocks and Shares ISA'. No web search fired. The same account's ChatGPT caught a different false premise (a stated £2,000 Personal Savings Allowance, versus the real £1,000) moments later where a search did fire, citing gov.uk. Claude, Gemini, Perplexity and Grok all caught the £25,000 error under identical default conditions.",
      "screenshot": null,
      "source_post": {
        "title": "AI ISA advice: I tested four tools on the questions people get wrong",
        "url": "https://dixon.ai/posts/ai-isa-accuracy-test/"
      }
    },
    {
      "id": "gemini-opaque-provenance-2026-07-07",
      "url": "https://dixon.ai/evidence/?outcome=wrong#gemini-opaque-provenance-2026-07-07",
      "tool": "gemini",
      "model_version": null,
      "error": "opaque-provenance",
      "hallucination_type": null,
      "consequence": "decision-affecting",
      "date": "2026-07-07",
      "prompt_run_date": null,
      "summary": "Asked for the maximum UK handheld-phone driving fine with a source, Gemini (Pro) put an opaque Police.uk label beside the correct £2,500 lorry-and-bus figure in two rounds and an opaque RAC label beside it in the third. The only inspectable receipt was the correct GOV.UK guide, detached at the bottom. The historical inline destinations cannot be recovered, so this is an auditability failure, not proof that Police.uk lacked the figure.",
      "screenshot": null,
      "source_post": {
        "title": "AI cites the wrong source: I put 6 UK questions to 5 assistants and checked every inspectable source",
        "url": "https://dixon.ai/posts/ai-cites-wrong-source/"
      }
    },
    {
      "id": "perplexity-stale-figure-as-current-2026-07-07",
      "url": "https://dixon.ai/evidence/?outcome=wrong#perplexity-stale-figure-as-current-2026-07-07",
      "tool": "perplexity",
      "model_version": null,
      "error": "stale-figure-as-current",
      "hallucination_type": null,
      "consequence": "decision-affecting",
      "date": "2026-07-07",
      "prompt_run_date": null,
      "summary": "Asked how many free childcare hours a working parent of a 9-month-old in England gets right now, Perplexity said 15 hours and described the 30-hour rollout as still to come, ten months after it completed. It cited a real Feb-2025 gov.uk page, and another of its own cited gov.uk sources states the opposite. On a same-day re-run it self-corrected to the right 30 hours, so the failure is intermittent, not fixed.",
      "screenshot": null,
      "source_post": {
        "title": "AI cites the wrong source: I put 6 UK questions to 5 assistants and checked every inspectable source",
        "url": "https://dixon.ai/posts/ai-cites-wrong-source/"
      }
    },
    {
      "id": "perplexity-misattributed-source-2026-07-07",
      "url": "https://dixon.ai/evidence/?outcome=wrong#perplexity-misattributed-source-2026-07-07",
      "tool": "perplexity",
      "model_version": null,
      "error": "misattributed-source",
      "hallucination_type": null,
      "consequence": "decision-affecting",
      "date": "2026-07-07",
      "prompt_run_date": null,
      "summary": "Asked for the maximum UK handheld-phone driving fine with a source, Perplexity led the £1,000 court fine with a solicitors'-firm marketing page over the gov.uk guide, citing that page four times inline in one round. The gov.uk link it did carry was an older press release, not the canonical guide. Unlike the childcare miss, this sourcing miss held across all three rounds.",
      "screenshot": null,
      "source_post": {
        "title": "AI cites the wrong source: I put 6 UK questions to 5 assistants and checked every inspectable source",
        "url": "https://dixon.ai/posts/ai-cites-wrong-source/"
      }
    },
    {
      "id": "claude-misattributed-source-2026-07-07",
      "url": "https://dixon.ai/evidence/?outcome=wrong#claude-misattributed-source-2026-07-07",
      "tool": "claude",
      "model_version": null,
      "error": "misattributed-source",
      "hallucination_type": null,
      "consequence": "cosmetic-but-revealing",
      "date": "2026-07-07",
      "prompt_run_date": null,
      "summary": "Gave the correct £1,000 and £2,500 court fines for using a handheld phone while driving, but sourced them to a solicitor firm's page rather than the gov.uk page that carries all three figures (which it cited separately, only for the £200 fixed penalty). Right numbers, wrong-tier citation for the figure the reader most wants.",
      "screenshot": null,
      "source_post": {
        "title": "AI cites the wrong source: I put 6 UK questions to 5 assistants and checked every inspectable source",
        "url": "https://dixon.ai/posts/ai-cites-wrong-source/"
      }
    },
    {
      "id": "gemini-incomplete-jurisdiction-2026-07-07",
      "url": "https://dixon.ai/evidence/?outcome=wrong#gemini-incomplete-jurisdiction-2026-07-07",
      "tool": "gemini",
      "model_version": null,
      "error": "incomplete-jurisdiction",
      "hallucination_type": null,
      "consequence": "decision-affecting",
      "date": "2026-07-07",
      "prompt_run_date": null,
      "summary": "Asked for stamp duty on a £300,000 home with the official page, Gemini gave the correct £5,000 England and Northern Ireland figure on the correct gov.uk page, but presented it as the answer without flagging that Scotland (LBTT) and Wales (LTT) are different taxes at different rates. ChatGPT and Grok both flagged the divergence unprompted.",
      "screenshot": null,
      "source_post": {
        "title": "AI cites the wrong source: I put 6 UK questions to 5 assistants and checked every inspectable source",
        "url": "https://dixon.ai/posts/ai-cites-wrong-source/"
      }
    },
    {
      "id": "gemini-misattribution-2026-07-07",
      "url": "https://dixon.ai/evidence/?outcome=wrong#gemini-misattribution-2026-07-07",
      "tool": "gemini",
      "model_version": null,
      "error": "misattribution",
      "hallucination_type": "faithfulness",
      "consequence": "decision-affecting",
      "date": "2026-07-07",
      "prompt_run_date": null,
      "summary": "Gave the correct £2,500 maximum fine for using a handheld phone while driving, but attached it to unresolvable Police.uk source chips in two rounds and an unresolvable RAC chip in the third; the inspectable gov.uk link sat separately at the bottom.",
      "screenshot": null,
      "source_post": {
        "title": "How accurate is Google Gemini? I graded 27 of its answers",
        "url": "https://dixon.ai/posts/how-accurate-is-google-gemini/"
      }
    },
    {
      "id": "perplexity-private-mode-identity-leak-2026-07-06",
      "url": "https://dixon.ai/evidence/?outcome=wrong#perplexity-private-mode-identity-leak-2026-07-06",
      "tool": "perplexity",
      "model_version": "Perplexity Incognito (logged in)",
      "error": "private-mode-identity-leak",
      "hallucination_type": "faithfulness",
      "consequence": "cosmetic-but-revealing",
      "date": "2026-07-06",
      "prompt_run_date": "2026-07-06",
      "summary": "Asked a generic question with zero personal context, 'Should I buy a house now, or keep renting and invest the difference? Give me a clear recommendation.', in Perplexity's Incognito mode while logged in, all three runs placed me in my region, the nearest big city, about 30 miles from where I actually live, and two of the three also greeted me by my real first name. One example heading it produced: 'Recommendation for you (first name, nearest city)'. The name is exact and comes from the account, because Incognito does not log you out. The location is approximate: it lands on the nearest big city rather than my actual town, which suggests it comes from something coarse like my internet address rather than anything I typed, though I did not run a control to prove whether it is the connection or a stored profile field. Perplexity's Incognito only promises the chat 'won't save to your history and expire after 24 hours', it does not promise anonymity, so this is not a broken promise. It is the gap between what 'incognito' implies and what the mode actually does. By contrast, Claude's Incognito chat named that same nearest city twice across eight runs in two separate sittings the same day, and no name; ChatGPT's Temporary Chat asked for my country and city rather than assuming, zero of three, though it ran no web search in any run while Perplexity searched every time.",
      "screenshot": null,
      "source_post": {
        "title": "I opened a private AI chat. It still knew my name and my rough location.",
        "url": "https://dixon.ai/posts/ai-knew-things-i-never-told-it/"
      }
    },
    {
      "id": "chatgpt-reversed-correct-answer-under-pushback-2026-07-05",
      "url": "https://dixon.ai/evidence/?outcome=wrong#chatgpt-reversed-correct-answer-under-pushback-2026-07-05",
      "tool": "chatgpt",
      "model_version": "ChatGPT (Free)",
      "error": "reversed-correct-answer-under-pushback",
      "hallucination_type": null,
      "consequence": "decision-affecting",
      "date": "2026-07-05",
      "prompt_run_date": "2026-07-05",
      "summary": "Asked a global tracker fund's yearly charge, ChatGPT gave the correct 0.19% at first. Pushed back with 'no, it's 0.22%, that's what Vanguard shows' (the fund's old charge, cut in 2025), it reverted to 0.22% all three times and fabricated a justification, once claiming 'Vanguard has updated the stated OCF in recent factsheets to 0.22%, which is the most reliable source' (false, the factsheets at the time said 0.19%). It took the source I'd named on trust rather than re-checking the page.",
      "screenshot": null,
      "source_post": {
        "title": "Does AI change its answer when you push back? I told five AIs they were wrong",
        "url": "https://dixon.ai/posts/does-ai-cave-when-you-push-back/"
      }
    },
    {
      "id": "grok-unit-denomination-2026-06-28",
      "url": "https://dixon.ai/evidence/?outcome=wrong#grok-unit-denomination-2026-06-28",
      "tool": "grok",
      "model_version": "Grok Fast (free tier, memory off, web search on)",
      "error": "unit-denomination",
      "hallucination_type": "factuality",
      "consequence": "decision-affecting",
      "date": "2026-06-28",
      "prompt_run_date": "2026-06-28",
      "summary": "Grok (free, 'Fast'): asked for BitMine Immersion's (BMNR) most recent full-year revenue, it returned '$6,095' (about $6K) on one run of three, instead of the correct $6.095 million from the SEC filing (a US company's annual report). Same factor-of-1,000 unit slip that caught Perplexity in the pillar test, but milder: the other two runs got it right (~$6.1M). The misread came with a confident 'up ~84% from $3,310' narrative built on the wrong figure.",
      "screenshot": "https://dixon.ai/screenshots/is-grok-good-for-stock-research/grok-d1-unit-error.png",
      "source_post": {
        "title": "Is Grok good for stock research? What four tests showed",
        "url": "https://dixon.ai/posts/is-grok-good-for-stock-research/"
      }
    },
    {
      "id": "perplexity-internal-contradiction-wrong-lead-2026-06-26",
      "url": "https://dixon.ai/evidence/?outcome=wrong#perplexity-internal-contradiction-wrong-lead-2026-06-26",
      "tool": "perplexity",
      "model_version": null,
      "error": "internal-contradiction-wrong-lead",
      "hallucination_type": "faithfulness",
      "consequence": "decision-affecting",
      "date": "2026-06-26",
      "prompt_run_date": "2026-06-26",
      "summary": "Asked whether a kettle faulty after three weeks qualifies for a full refund, one run of three opened by saying three weeks fell after the 30-day short-term right to reject, then corrected itself two paragraphs later. 21 days is inside 30. The law cited was correct throughout; only the order was wrong.",
      "screenshot": null,
      "source_post": {
        "title": "An AI told me three weeks was more than thirty days",
        "url": "https://dixon.ai/posts/ai-uk-refund-rights-kettle/"
      }
    },
    {
      "id": "other-satire-logged-as-fact-2026-06-26",
      "url": "https://dixon.ai/evidence/?outcome=wrong#other-satire-logged-as-fact-2026-06-26",
      "tool": "other",
      "model_version": null,
      "error": "satire-logged-as-fact",
      "hallucination_type": null,
      "consequence": "cosmetic-but-revealing",
      "date": "2026-06-26",
      "prompt_run_date": "2026-06-26",
      "summary": "The site's own automated news-radar, sweeping for AI-reliability stories on 26 June 2026, logged Andrew Nesbitt's satirical 'Incident Report: CVE-2026-LGTM' (published the same day, with a 'satire' topic tag directly beneath the title) as, verbatim, 'a real production AI reliability failure. Documented, citable, primary source available,' and routed it as a post candidate. Caught before publication by reading the primary source, which carried the satire tag in plain sight.",
      "screenshot": "https://dixon.ai/screenshots/my-ai-radar-believed-a-fake/satire-tag.png",
      "source_post": {
        "title": "I run an AI to catch AI mistakes. It fell for a fake.",
        "url": "https://dixon.ai/posts/my-ai-radar-believed-a-fake/"
      }
    },
    {
      "id": "chatgpt-misattributed-source-2026-06-20",
      "url": "https://dixon.ai/evidence/?outcome=wrong#chatgpt-misattributed-source-2026-06-20",
      "tool": "chatgpt",
      "model_version": null,
      "error": "misattributed-source",
      "hallucination_type": null,
      "consequence": "decision-affecting",
      "date": "2026-06-20",
      "prompt_run_date": null,
      "summary": "Asked for the UK ISA partial-transfer rule with a source, ChatGPT (free, web search on) cited gov.uk/individual-savings-accounts/if-you-move-abroad-or-die, a real, live gov.uk page about what happens to an ISA when you move abroad or die. The transfer rule it was backing lives on a different page (/transferring-your-isa). The URL resolved; it just didn't hold the claim.",
      "screenshot": "https://dixon.ai/screenshots/can-you-trust-chatgpt-sources/chatgpt-isa-wrong-page.png",
      "source_post": {
        "title": "Does ChatGPT make up sources? I checked two finance claims against the actual pages",
        "url": "https://dixon.ai/posts/can-you-trust-chatgpt-sources/"
      }
    },
    {
      "id": "perplexity-low-authority-source-led-2026-06-20",
      "url": "https://dixon.ai/evidence/?outcome=wrong#perplexity-low-authority-source-led-2026-06-20",
      "tool": "perplexity",
      "model_version": "Best (auto-routing, underlying model unknown)",
      "error": "low-authority-source-led",
      "hallucination_type": null,
      "consequence": "cosmetic-but-revealing",
      "date": "2026-06-20",
      "prompt_run_date": "2026-06-20",
      "summary": "Asked how long cooked chicken keeps in the fridge by a stated UK (Newcastle) user, Perplexity (web search on) led with US food blogs, Martha Stewart, Springer Mountain Farms, and gave the US figure of 3-4 days. The UK FSA guidance (2 days for cooked leftovers, per food.gov.uk) appeared as a secondary note, not the primary answer. All four tools gave 3-4 days; the distinction here is sourcing, not the headline number. Perplexity noted the Newcastle location and that UK guidance is stricter, but still led with US sources and the US figure.",
      "screenshot": null,
      "source_post": {
        "title": "Does web search make AI more accurate? I ran the same questions both ways",
        "url": "https://dixon.ai/posts/web-search-makes-ai-differently-unreliable/"
      }
    },
    {
      "id": "perplexity-outdated-rule-stated-as-current-2026-06-19",
      "url": "https://dixon.ai/evidence/?outcome=wrong#perplexity-outdated-rule-stated-as-current-2026-06-19",
      "tool": "perplexity",
      "model_version": "Best (auto-routing, underlying model unknown)",
      "error": "outdated-rule-stated-as-current",
      "hallucination_type": null,
      "consequence": "decision-affecting",
      "date": "2026-06-19",
      "prompt_run_date": "2026-06-19",
      "summary": "Asked whether this year's ISA contributions can be partially transferred, Perplexity said they must be transferred in full, the rule abolished on 6 April 2024. Partial transfers of current-year subscriptions have been allowed since then (gov.uk). Stated with no date and no hedge. ChatGPT (Free) gave the same outdated answer.",
      "screenshot": null,
      "source_post": {
        "title": "AI ISA advice: I tested four tools on the questions people get wrong",
        "url": "https://dixon.ai/posts/ai-isa-accuracy-test/"
      }
    },
    {
      "id": "chatgpt-outdated-rule-stated-as-current-2026-06-19",
      "url": "https://dixon.ai/evidence/?outcome=wrong#chatgpt-outdated-rule-stated-as-current-2026-06-19",
      "tool": "chatgpt",
      "model_version": "ChatGPT Free",
      "error": "outdated-rule-stated-as-current",
      "hallucination_type": null,
      "consequence": "decision-affecting",
      "date": "2026-06-19",
      "prompt_run_date": "2026-06-19",
      "summary": "Same miss as Perplexity: stated the pre-6-April-2024 'transfer current-year ISA money in full' rule as if current, no date, no search. Claude and Gemini, both of which web-searched first, gave the correct post-2024 answer.",
      "screenshot": null,
      "source_post": {
        "title": "AI ISA advice: I tested four tools on the questions people get wrong",
        "url": "https://dixon.ai/posts/ai-isa-accuracy-test/"
      }
    },
    {
      "id": "chatgpt-unsupported-premium-estimates-2026-06-18",
      "url": "https://dixon.ai/evidence/?outcome=wrong#chatgpt-unsupported-premium-estimates-2026-06-18",
      "tool": "chatgpt",
      "model_version": "ChatGPT Free plan, unnamed rate-limited fallback, web search on",
      "error": "unsupported-premium-estimates",
      "hallucination_type": "factuality",
      "consequence": "decision-affecting",
      "date": "2026-06-18",
      "prompt_run_date": "2026-06-18",
      "summary": "Asked for AAPL covered-call strikes and premiums with no chain data supplied, ChatGPT generated a hypothetical premium table with specific dollar ranges and yields, an assumed 25% implied volatility, and a generic Barchart citation. It explicitly said the ranges were not live quotes, but opened by claiming to use 'the latest available options-chain data' and never identified a reproducible chain snapshot. The UI called this an unnamed 'less powerful model' after the Free-plan limit was reached; the exact model was not shown.",
      "screenshot": "https://dixon.ai/screenshots/ai-stock-research-tools-tested/chatgpt-options-premium-table-2026-06-18-receipt.webp",
      "source_post": {
        "title": "AI stock research tools tested: 3 failed, 1 stayed clean",
        "url": "https://dixon.ai/posts/ai-stock-research-tools-tested/"
      }
    },
    {
      "id": "perplexity-inconsistent-5yr-returns-2026-06-13",
      "url": "https://dixon.ai/evidence/?outcome=wrong#perplexity-inconsistent-5yr-returns-2026-06-13",
      "tool": "perplexity",
      "model_version": null,
      "error": "inconsistent-5yr-returns",
      "hallucination_type": "factuality",
      "consequence": null,
      "date": "2026-06-13",
      "prompt_run_date": null,
      "summary": "Tabled two 5-year returns from different sources side by side without units (VWRL 11.83% next to VUSA 86.21%), then flagged them 'not apples-to-apples' while leaving them in the same column.",
      "screenshot": null,
      "source_post": {
        "title": "Is ChatGPT accurate? I asked four AIs one simple money question and checked every number",
        "url": "https://dixon.ai/posts/is-chatgpt-accurate/"
      }
    },
    {
      "id": "gemini-unprompted-cross-conversation-memory-2026-06-13",
      "url": "https://dixon.ai/evidence/?outcome=wrong#gemini-unprompted-cross-conversation-memory-2026-06-13",
      "tool": "gemini",
      "model_version": null,
      "error": "unprompted-cross-conversation-memory",
      "hallucination_type": "faithfulness",
      "consequence": null,
      "date": "2026-06-13",
      "prompt_run_date": null,
      "summary": "Injected personal context from earlier chats into a standard fund comparison, unprompted. This account's answer used prior-chat context absent from the prompt, so the response was not reproducible from the visible question alone.",
      "screenshot": null,
      "source_post": {
        "title": "Is ChatGPT accurate? I asked four AIs one simple money question and checked every number",
        "url": "https://dixon.ai/posts/is-chatgpt-accurate/"
      }
    },
    {
      "id": "claude-stale-figure-with-web-search-2026-06-13",
      "url": "https://dixon.ai/evidence/?outcome=wrong#claude-stale-figure-with-web-search-2026-06-13",
      "tool": "claude",
      "model_version": null,
      "error": "stale-figure-with-web-search",
      "hallucination_type": "factuality",
      "consequence": null,
      "date": "2026-06-13",
      "prompt_run_date": null,
      "summary": "Served the out-of-date 0.22% ongoing charge for VWRL despite running a web search before answering; the published figure at the time was 0.19%.",
      "screenshot": null,
      "source_post": {
        "title": "Is ChatGPT accurate? I asked four AIs one simple money question and checked every number",
        "url": "https://dixon.ai/posts/is-chatgpt-accurate/"
      }
    },
    {
      "id": "gemini-fabricated-interface-element-2026-06-12",
      "url": "https://dixon.ai/evidence/?outcome=wrong#gemini-fabricated-interface-element-2026-06-12",
      "tool": "gemini",
      "model_version": "Gemini (Pro)",
      "error": "fabricated-interface-element",
      "hallucination_type": "factuality",
      "consequence": "cosmetic-but-revealing",
      "date": "2026-06-12",
      "prompt_run_date": "2026-06-12",
      "summary": "Asked 'should I buy NVDA?' in a fresh session on 12 June 2026 (web search on), Gemini's stored response text ended with 'Asset Record Saved: NVIDIA Corporation (NVDA) has been logged with its Q1 FY27 financial details' and printed 'Evaluate options for covered calls? Yes'. No corresponding record or working control appeared outside the answer. This finding is text-capture evidence: the published session screenshot shows an earlier part of the response, not those lines.",
      "screenshot": null,
      "source_post": {
        "title": "AI stock picker: I asked three models if I should buy NVDA, and watched the methodology break",
        "url": "https://dixon.ai/posts/ai-stock-picker/"
      }
    },
    {
      "id": "gemini-vague-source-attribution-2026-06-12",
      "url": "https://dixon.ai/evidence/?outcome=wrong#gemini-vague-source-attribution-2026-06-12",
      "tool": "gemini",
      "model_version": "Gemini (Pro)",
      "error": "vague-source-attribution",
      "hallucination_type": null,
      "consequence": "cosmetic-but-revealing",
      "date": "2026-06-12",
      "prompt_run_date": "2026-06-12",
      "summary": "Asked where its trailing P/E of 30.69 came from, Gemini attributed the precise figure to 'standard retail financial data platforms, such as Yahoo Finance and Robinhood' with no specific source and no link, a gesture at the kind of place such a number might live rather than a checkable citation.",
      "screenshot": "https://dixon.ai/screenshots/ai-stock-picker/gemini-nvda-provenance.webp",
      "source_post": {
        "title": "AI stock picker: I asked three models if I should buy NVDA, and watched the methodology break",
        "url": "https://dixon.ai/posts/ai-stock-picker/"
      }
    },
    {
      "id": "chatgpt-fabricated-live-price-2026-06-11",
      "url": "https://dixon.ai/evidence/?outcome=wrong#chatgpt-fabricated-live-price-2026-06-11",
      "tool": "chatgpt",
      "model_version": null,
      "error": "fabricated-live-price",
      "hallucination_type": "factuality",
      "consequence": null,
      "date": "2026-06-11",
      "prompt_run_date": null,
      "summary": "Asked for NVDA's current share price in two fresh sessions on 11 June 2026, ChatGPT gave $206.18 'live' (NVDA's real high that day was $205.66, so that figure never printed) and, in the second run, $191.21 'during today's session', which was $8.33 below the real day's low of $199.54. Neither price existed at any point that day; both were presented with citations.",
      "screenshot": null,
      "source_post": {
        "title": "How often is ChatGPT wrong? I kept a running tally across 20 real AI tests",
        "url": "https://dixon.ai/posts/how-often-is-chatgpt-wrong/"
      }
    },
    {
      "id": "gemini-wrong-entity-audit-2026-06-10",
      "url": "https://dixon.ai/evidence/?outcome=wrong#gemini-wrong-entity-audit-2026-06-10",
      "tool": "gemini",
      "model_version": "Gemini Flash",
      "error": "wrong-entity-audit",
      "hallucination_type": null,
      "consequence": "decision-affecting",
      "date": "2026-06-10",
      "prompt_run_date": null,
      "summary": "Asked to review 'Dixon Dixon AI' (a voice-input transcription of dixon.ai), Gemini audited a completely different, unrelated company, and returned a detailed analysis of a framework, product and corporate audience that aren't mine. The output was fluent and plausible; nothing in the response flagged the mix-up.",
      "screenshot": "https://dixon.ai/screenshots/gemini-audited-the-wrong-business/gemini-audit-02-wrong-business.webp",
      "source_post": {
        "title": "Gemini audited my website, and reviewed a different business entirely",
        "url": "https://dixon.ai/posts/gemini-audited-the-wrong-business/"
      }
    },
    {
      "id": "gemini-stale-memory-as-current-2026-06-10",
      "url": "https://dixon.ai/evidence/?outcome=wrong#gemini-stale-memory-as-current-2026-06-10",
      "tool": "gemini",
      "model_version": "Gemini Flash",
      "error": "stale-memory-as-current",
      "hallucination_type": null,
      "consequence": "cosmetic-but-revealing",
      "date": "2026-06-10",
      "prompt_run_date": null,
      "summary": "In a second session naming dixon.ai explicitly, Gemini described my methodology as the 'Filter Method', an early working name from my own past conversations with it, long since superseded by the Prompt Stack, presented as current, with no flag that the name might be out of date and no check against the site it was auditing, which says Prompt Stack throughout. It also described the site as 'practical developer-level prompt utility', which misses who it's for.",
      "screenshot": "https://dixon.ai/screenshots/gemini-audited-the-wrong-business/gemini-audit-seo-conv-01-prompt.webp",
      "source_post": {
        "title": "Gemini audited my website, and reviewed a different business entirely",
        "url": "https://dixon.ai/posts/gemini-audited-the-wrong-business/"
      }
    },
    {
      "id": "gemini-partial-fabrication-2026-05-22",
      "url": "https://dixon.ai/evidence/?outcome=wrong#gemini-partial-fabrication-2026-05-22",
      "tool": "gemini",
      "model_version": "Gemini Pro",
      "error": "partial-fabrication",
      "hallucination_type": null,
      "consequence": "decision-affecting",
      "date": "2026-05-22",
      "prompt_run_date": "2026-05-22",
      "summary": "Re-ran the BMNR covered-call no-chain test from 2026-05-15. Gemini correctly listed three data points needing a live chain (bid/ask spreads, precise delta and premium output), then in the same response supplied a 75-90% IV expectation and a 20-30 delta range for a 15% OTM 45-day strike without a live chain or cited source. The result supports an internal provenance contradiction, not proof that either range was numerically false.",
      "screenshot": null,
      "source_post": {
        "title": "What AI stock research comparisons should test",
        "url": "https://dixon.ai/posts/comparison-articles-get-wrong/"
      }
    },
    {
      "id": "claude-stale-prompt-framing-2026-05-20",
      "url": "https://dixon.ai/evidence/?outcome=wrong#claude-stale-prompt-framing-2026-05-20",
      "tool": "claude",
      "model_version": "Claude Opus 4.7",
      "error": "stale-prompt-framing",
      "hallucination_type": null,
      "consequence": "cosmetic-but-revealing",
      "date": "2026-05-20",
      "prompt_run_date": "2026-05-20",
      "summary": "Re-ran two prompts on Claude Opus 4.7 with live search on. Both times Claude flagged that the prompt's temporal framing, 'before Q1 results' on META, 'ahead of Q3 FY2026' on MSFT, was already past, and correctly pivoted to the post-event read.",
      "screenshot": "https://dixon.ai/screenshots/claude-prompts-for-investing-real-outputs/claude-meta-bull-bear.webp",
      "source_post": {
        "title": "Claude prompts for investing: 6 real examples",
        "url": "https://dixon.ai/posts/claude-prompts-for-investing-real-outputs/"
      }
    },
    {
      "id": "gemini-unsupported-premium-estimates-2026-05-16",
      "url": "https://dixon.ai/evidence/?outcome=wrong#gemini-unsupported-premium-estimates-2026-05-16",
      "tool": "gemini",
      "model_version": "Gemini 2.5 Pro",
      "error": "unsupported-premium-estimates",
      "hallucination_type": null,
      "consequence": "decision-affecting",
      "date": "2026-05-16",
      "prompt_run_date": "2026-05-16",
      "summary": "Given only BMNR's share price, supplied current-looking IV, IV Rank, strikes and premium estimates while claiming they were based on 'current order book data'. The preserved unconnected session contains no source or broker comparison supporting that provenance claim, so the figures were not safe to use as live quotes.",
      "screenshot": "https://dixon.ai/screenshots/ai-limitations-options-trading/gemini-chain-fabrication.webp",
      "source_post": {
        "title": "AI for options trading: 4 workflows and 6 data guardrails",
        "url": "https://dixon.ai/posts/ai-limitations-options-trading/"
      }
    },
    {
      "id": "claude-inferred-input-2026-05-16",
      "url": "https://dixon.ai/evidence/?outcome=wrong#claude-inferred-input-2026-05-16",
      "tool": "claude",
      "model_version": "Claude Opus 4.7 (Max)",
      "error": "inferred-input",
      "hallucination_type": null,
      "consequence": "decision-affecting",
      "date": "2026-05-16",
      "prompt_run_date": "2026-05-16",
      "summary": "Estimated a BMNR $23 call's probability of finishing in the money using Black-Scholes N(d2) and a rough 90–110% volatility range derived from web references rather than the live contract. Claude disclosed the estimates, returned ranges and told Ben to check the broker; the result was transparent but too input-sensitive to trade on.",
      "screenshot": "https://dixon.ai/screenshots/ai-limitations-options-trading/claude-w6-assignment-probability.webp",
      "source_post": {
        "title": "AI for options trading: 4 workflows and 6 data guardrails",
        "url": "https://dixon.ai/posts/ai-limitations-options-trading/"
      }
    },
    {
      "id": "perplexity-ignored-constraint-2026-05-15",
      "url": "https://dixon.ai/evidence/?outcome=wrong#perplexity-ignored-constraint-2026-05-15",
      "tool": "perplexity",
      "model_version": "Perplexity Pro (default)",
      "error": "ignored-constraint",
      "hallucination_type": null,
      "consequence": "cosmetic-but-revealing",
      "date": "2026-05-15",
      "prompt_run_date": "2026-05-15",
      "summary": "On a Meta Q1 2026 earnings prompt that explicitly instructed 'work only from the pasted document', Perplexity ran 10 external web searches. The output was technically correct but came from external coverage of the release rather than reasoning over the supplied transcript. Not a bug, Perplexity routes to search as its default behaviour, but a constraint-following failure that matters when the test is designed to measure document discipline. Same prompt run on ChatGPT and Claude stayed inside the document.",
      "screenshot": null,
      "source_post": {
        "title": "Best AI for Earnings Reports? ChatGPT vs Claude vs Perplexity",
        "url": "https://dixon.ai/posts/best-ai-tools-for-earnings-analysis/"
      }
    },
    {
      "id": "perplexity-unit-error-2026-05-14",
      "url": "https://dixon.ai/evidence/?outcome=wrong#perplexity-unit-error-2026-05-14",
      "tool": "perplexity",
      "model_version": "Perplexity Sonar Pro",
      "error": "unit-error",
      "hallucination_type": null,
      "consequence": "decision-affecting",
      "date": "2026-05-14",
      "prompt_run_date": "2026-05-14",
      "summary": "Read BMNR revenue as $6K instead of $6.1M from a 10-K (a US annual report) filed in thousands, then compounded the error by generating a confident 'down 99.8% from prior year' decline narrative around the wrong figure. A retail investor acting on this would have a materially false picture of the business. (Re-tested 18 June 2026: did not reproduce. Perplexity returned the correct ~$6.1M figure. Logged as a dated, point-in-time failure.)",
      "screenshot": "https://dixon.ai/screenshots/chatgpt-vs-claude-vs-perplexity-stock-research/perplexity-d1-unit-error.webp",
      "source_post": {
        "title": "Claude vs ChatGPT vs Gemini for stock analysis: who bluffed?",
        "url": "https://dixon.ai/posts/chatgpt-vs-claude-vs-perplexity-stock-research/"
      }
    },
    {
      "id": "gemini-fabrication-2026-05-14",
      "url": "https://dixon.ai/evidence/?outcome=wrong#gemini-fabrication-2026-05-14",
      "tool": "gemini",
      "model_version": "Gemini 2.5 Pro (deep thinking)",
      "error": "fabrication",
      "hallucination_type": null,
      "consequence": "decision-affecting",
      "date": "2026-05-14",
      "prompt_run_date": "2026-05-14",
      "summary": "Returned a formatted covered-call comparison table with specific premium estimates ($3.50–$4.00 for the $26 strike, etc.), made up an implied volatility figure of ~75%, used the wrong stock price ($28.60 vs $21.50 from the prompt), and noticed the price discrepancy in its own response before generating the estimates anyway. (Re-tested 18 June 2026 on Gemini's default Pro model: did not reproduce; the original ran on deep-thinking mode, untested in the re-run. Logged as a dated, point-in-time failure.)",
      "screenshot": "https://dixon.ai/screenshots/chatgpt-vs-claude-vs-perplexity-stock-research/gemini-d5-invented-premiums-receipt.webp",
      "source_post": {
        "title": "Claude vs ChatGPT vs Gemini for stock analysis: who bluffed?",
        "url": "https://dixon.ai/posts/chatgpt-vs-claude-vs-perplexity-stock-research/"
      }
    },
    {
      "id": "perplexity-unit-error-2026-05-14",
      "url": "https://dixon.ai/evidence/?outcome=wrong#perplexity-unit-error-2026-05-14",
      "tool": "perplexity",
      "model_version": null,
      "error": "unit-error",
      "hallucination_type": null,
      "consequence": "decision-affecting",
      "date": "2026-05-14",
      "prompt_run_date": null,
      "summary": "On BMNR, Perplexity read a 10-K reported 'in thousands' literally, turning $6,095 thousand ($6.1m) into '$6K', then narrated a confident 'down 99.8% from prior year' decline that never happened. The exact 18 June 2026 rerun returned the correct figure. The error and clean rerun are dated outcomes; these captures do not isolate company coverage as the cause.",
      "screenshot": "https://dixon.ai/screenshots/chatgpt-vs-claude-vs-perplexity-stock-research/perplexity-d1-unit-error.webp",
      "source_post": {
        "title": "Is Perplexity good for investment research? One 1,000× error, one clean rerun",
        "url": "https://dixon.ai/posts/is-perplexity-good-for-investment-research/"
      }
    }
  ]
}