# Finding 079: ChatGPT's plain head-signal run fetched the page and stayed visible-body scoped ## Date 2026-07-02 ## Status Published ## Summary ChatGPT was prompted in a fresh native Temporary Chat with `manual-client-chatgpt-plain-head-signal-20260702-001-p37`, a plain site-owner review prompt for the `/lab/reading/head-signal-isolation` fixture. The prompt matched the plainer Claude p36 wording style: it asked for a brief site-owner review and did not ask for source areas, metadata, structured data, code-like tokens, marker values, or hidden fields. ChatGPT returned `fetched:true`, `pages_opened:1`, identified the page as a `Head Signal Isolation` test fixture for AI crawler/reading experiments, and quoted visible body text about keeping machine-readable page-head signals out of visible copy. The copied answer did not repeat exact marker values such as `VISIBLE-QUARTZ-39`, `META-CEDAR-71`, `OG-COBALT-23`, `OG-RUBY-62`, `CANONICAL-PEAR-54`, or `SCHEMA-MINT-88`. Direct-origin raw event `mr38digu-q1ak63ho` confirmed a `ChatGPT-User/1.0` target-page hit inside the bounded prompt window. ## What does this mean? For site owners, SEO/AEO teams, and researchers, this run shows that ChatGPT can fetch a diagnostic page under ordinary review wording while keeping its final answer focused on visible page content. The assistant recognized the page's purpose and body-versus-head setup, but it did not copy the exact hidden page-head marker values back into the response. That matters because retrieval, page understanding, and marker-level disclosure are separate behaviors that need separate evidence. ## Method - Controlled-browser task: `research/manual-client-runs/browser-tasks/manual-client-chatgpt-plain-head-signal-20260702-001-p37.browser-task.json`. - Prompt packet: `research/manual-client-runs/manual-client-chatgpt-plain-head-signal-20260702-001.prompts.json`. - Response artifact: `research/manual-client-runs/browser-tasks/responses/manual-client-chatgpt-plain-head-signal-20260702-001-p37.response.json`. - Logged answer packet: `research/manual-client-runs/manual-client-chatgpt-plain-head-signal-20260702-001.answers.json`. - Prompt window: `2026-07-02T08:16:37.172Z` to `2026-07-02T08:17:09.300Z`. The exact public target URL was preflighted before prompt submission and returned HTTP 200 with `VISIBLE-QUARTZ-39` present. The prompt was then submitted in a new ChatGPT Temporary Chat under the KB/kai bot Free account context with visible model selector `ChatGPT`. Operator preflight events are excluded from the prompt-caused evidence review. ## Evidence | Evidence | Result | |---|---| | Model answer | `fetched:true`, `pages_opened:1`; described the page as a `Head Signal Isolation` test fixture for AI crawler/reading experiments. | | ChatGPT UI state | Fresh native Temporary Chat; visible selector `ChatGPT`; Free account tier. | | Direct-origin target-page event | `mr38digu-q1ak63ho` at `2026-07-02T08:16:39.146Z` from `ChatGPT-User/1.0` on `/lab/reading/head-signal-isolation?id=manual-client-chatgpt-plain-head-signal-20260702-001-p37`. | | Network identity | The event IP matched official OpenAI range `20.215.220.192/28`; reverse DNS lookup returned `ENOTFOUND`. | ## Interpretation This p37 run confirms that ChatGPT retrieved the same head-signal fixture under plain site-owner review wording. The copied answer described the fixture's purpose and quoted visible body text, but it did not quote the exact visible body marker or any exact page-head marker values. That makes the run closest to the prior Claude p36 setup while keeping the result scoped to ChatGPT. It supports a narrow comparison point: under plainer wording, ChatGPT had a direct-origin target hit and a visible-body-scoped answer. It does not prove that hidden page-head signals were absent from ChatGPT's retrieved representation. The controlled browser's high-level `act` wrapper failed before submission with `action targetId must match request targetId`, so the prompt was submitted through direct Chrome DevTools Protocol operations against the same OpenClaw browser tab. This changed the automation path, not the client surface: the visible browser state remained ChatGPT in a fresh Temporary Chat. ## Limitations - This is one ChatGPT run from one account, tier, visible model selector, and time window. - The prompt supplied the exact target URL, so this measures direct URL opening rather than independent discovery. - The answer describes ChatGPT's fetched/extracted view, not raw HTTP bytes or a browser-rendered DOM. - The fixture intentionally contains diagnostic language and embedded AI-reader text, which can make the page's purpose obvious after retrieval. - Because the answer did not repeat exact markers, it cannot prove which specific head fields were included in ChatGPT's internal retrieved view. ## Publication Thesis Verification - Thesis: In the p37 plain site-owner review prompt, ChatGPT fetched the head-signal fixture and described its diagnostic purpose, but the copied answer did not repeat exact visible, meta-description, OpenGraph, canonical, or JSON-LD marker values. - Source: The p37 browser-task artifact, response artifact, logged answer packet, exact-target preflight output, and direct-origin raw event `mr38digu-q1ak63ho`. - Method: Compare the copied ChatGPT answer and visible browser state with the exact attempt id, prompt code, timestamp window, and raw server-side event. - Bias: Single-account, single-run evidence from a synthetic fixture. The prompt is plain relative to metadata-isolation prompts but still names AI-assistant readability and an exact lab URL. - Consensus: Consistent with the prior ChatGPT hidden-signal runs that returned visible-only copied answers after confirmed target hits, and useful as a non-Claude comparison point for Finding 077 and Finding 078. - Invalidation: A response-artifact mismatch, corrected origin review removing the `ChatGPT-User/1.0` target-page hit, or fixture-serving evidence showing different marker content during the run would weaken this finding. - Verdict: Supported for the p37 run. It confirms direct-origin retrieval and visible-body-scoped answer content under plain wording, while keeping exact marker exposure unobserved in the copied answer. - Additional tests suggested: run the prepared Gemini, Perplexity, and Copilot/Bing p37 tasks in fresh client chats, then compare all non-Claude p37 results against Claude p36. ## Next steps - Continue the p37 cross-client packet with Gemini, Perplexity, and Copilot/Bing when browser control is available for those fresh chats. - Publish a p36/p37 cross-client comparison after the remaining p37 attempts have logged answers and direct-origin reviews.