AI visibility
Brand citations on ChatGPT: the Pico Neural method
How to use Pico Neural's AI visibility page to connect your reading of a site page with brand citations observed in ChatGPT, Perplexity, Gemini, Grok and Google AI answers.
To use Pico Neural's AI visibility page, we suggest a four-step sequence: pick a page, record the citation you observed, compare the page topic with the citation topic, then check which engines you have connected. This is a Pico Neural reading proposal, not an automatic linking feature. AI visibility does not by itself connect your reading of site pages with observed citations. You note the brand citation by hand, next to the page you are reading, and mark the result yes, no or partial. Our public pages name ChatGPT, Perplexity, Gemini, Grok and Google AI. Claude and Grok are measured only after you connect them. Gemini and Google AI Overviews are not the same observation.
The Pico Neural method in four steps
The AI visibility page is where you read site pages and keep track of citations you have observed. Joining the two is annotation work we do ourselves, not a product output. Here are the steps, each with a checkable criterion.
- Pick the site page to read. Open a public page of the site in AI visibility and note the exact URL and the page topic in one sentence. Checkable criterion: the sentence describes the page's main subject, not the menu title or the product name.
- Record the citation you observed. For every ChatGPT, Perplexity, Gemini, Grok or Google AI answer that mentions the brand, note the engine, the date of the observation and the quoted text. Checkable criterion: the citation is copied word for word, with no paraphrase.
- Compare the page topic with the citation topic. Write the page topic (what the site page is about) and the citation topic (what the engine's answer is about) in two separate columns. Checkable criterion: the two columns hold distinct sentences, not the same sentence copied twice.
- Assign the result and check connected engines. Mark yes if the citation matches the page topic, no if it does not, partial if it matches only in part. Checkable criterion: for every partial result, note which part of the page is missing from the citation. Then check which engines you have connected: Claude and Grok are measured only after you connect them.
A filled-in row
Say a site page covers how to choose an energy supplier, and a ChatGPT answer cites the brand in connection with activation costs. In the page topic column we write "criteria for choosing an energy supplier"; in the citation topic column we write "supplier activation costs". The result is partial: the citation touches one aspect of the page, but the part on selection criteria is missing. We note the missing part, the selection criteria, in the same row, so the row stays checkable later.
What this method adds is a separation between two readings that are often confused: the reading of the site page and the reading of the observed citation. In AI visibility we read the page and sum it up as a topic in one sentence; outside AI visibility, in the engine's answer, we read the citation and sum it up as another topic. The link is not data the product produces; it is a comparison we make row by row. That is why the "page topic" column must describe the page's subject, not the brand or the engine, while the "citation topic" column must describe the answer's subject, not the page. Keeping the two columns distinct is what makes the yes, no or partial result checkable by anyone who picks up the sheet. If the two columns overlap, the result loses meaning and the comparison becomes a repetition. Pico Neural's working proposal is therefore to fill in one row per observed citation, with page, citation, page topic, citation topic and result, and not to edit a row once it is closed: if a correction is needed, add a new row with the date. That way the page reading and the citation stay two distinct elements, and the link can be reconstructed.
Limits to keep in mind
The method produces no rankings and no forecasts: it records only what was observed in an answer, on a date, on an engine. An API answer is not a person's screen, so the observation must be noted with the context in which it happened. Gemini and Google AI Overviews are not the same observation and belong in separate rows. Claude and Grok are measured only after you connect them, so a row that names them without a connection is not checkable. The yes, no or partial result criterion is a Pico Neural working proposal, not a universal standard: anyone picking up the sheet must be able to read the missing part noted for each partial result.
For the full picture on the AI visibility page and the engines named here, see the in-depth page.