Go-to-market

Chris Green shows how to test whether ChatGPT retrieves your page, and open-sourced a Chrome extension that automates the check

September 18, 2026 at 4:27 PM PT

Search Engine Journal illustration for checking retrieval pipeline for AI

Image: Search Engine Journal

Why it mattersA team that publishes for AI answers has had no equivalent of Search Console for ChatGPT or Perplexity. This test is a practical stand-in that tells you whether retrieval is the problem before you rewrite the page.

Chris Green has published a method for checking whether an AI chatbot can retrieve a specific page from its search source, and released a Chrome extension called Exactly Matchy that automates the check. The article ran in Search Engine Journal on 18 September and is a syndicated version of the Substack post at chrisgreenseo.substack.com. Google Search Console shows you what is in Google's index; there is nothing equivalent for ChatGPT or Perplexity, and this fills part of the gap.

The manual test

The idea copies the classical SEO trick of pasting a page's exact sentence into Google in quotes to see whether it is indexed. Instead of a search bar, you paste a snippet into a chatbot with a prompt like: "Search for '...' and return any results which contain that exact text only." Green ran this against ChatGPT signed out and got the URL back, which proved the retrieval pipeline reached the page.

If the URL does not come back, Green lists the questions to run in order: is the page discoverable through internal links or a sitemap, is it fetchable and not blocked by a WAF or robots.txt, is it crawlable and rendered, is it indexable, and is the snippet distinctive enough to be the only match. ChatGPT calls from more than one search source, so Green suggests testing four or five times with different snippets before drawing conclusions. The output is not truth, only an inference about what the assistant's index contains.

Exactly Matchy: on-device snippet picker

Exactly Matchy is a browser extension that Green open-sourced on GitHub. It reads the rendered page, strips navigation, cookie banners, and footers, and picks distinctive passages of 20 to 30 words that carry names, numbers, or uncommon wording rather than generic marketing copy. Chrome's on-device LLM ranks the candidates. Each picked passage becomes an exact-match prompt with one-click links into ChatGPT, Claude, and Gemini.

Green built the extension deliberately on-device so page text does not travel to a third-party API. There are no API keys, no per-check cost, and if Chrome AI is unavailable the extension falls back to a scoring method that runs without a model. The code lives at github.com/chr156r33n/exactly-matchy, and Green ships it as an unpacked developer-mode extension rather than a Chrome Web Store listing.

The workflow this gives an SEO team

The test does not replace Search Console; it does two things Search Console cannot. It confirms whether a specific chatbot can reach a specific URL, and, when it cannot, it separates a retrieval fault from a content one. If the snippet is distinctive and the URL does not return, the fix is a technical SEO fix: discovery, robots and firewall access, indexability, and crawlability. If the snippet is generic, the fix is content, because the passage is not strong enough to be the only match Google or Bing would return for the assistant to cite.

Green tells readers to review the extension's code before installing it, since a developer-mode extension from an unfamiliar source can read every page the browser opens.

Source

Reported by: Search Engine Journal

This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.

More from AI News

Zeeshan Yaseen ran two GEO experiments, and third-party listicles produced 85.8 percent of AI citations

Two hand-tracked GEO experiments by consultant Zeeshan Yaseen found third-party listicles produced 85.8 percent of AI citations, with owned listicles at 14 percent.

Source: PressGo-to-market

Carolyn Shelby in Search Engine Journal says AI search returns outdated brand facts because old pages still match the question

In a Search Engine Journal column published September 7, principal consultant Carolyn Shelby says brands get wrong answers in AI search because their own sites contain too many conflicting versions of the facts, and that publishing more content will not fix it.

Source: PressGo-to-market

CallRail says AI citation clicks now account for 1 to 2 percent of tracked local calls, and after-hours share of AI traffic reaches nearly two thirds

CallRail told Search Engine Journal that clicks from AI answer engines have roughly doubled since January and are now around 1 to 2 percent of tracked calls, and that nearly two thirds of visits from AI referrers happen outside business hours.

Source: PressGo-to-market