Answers
Pico Neural FAQ: what an AI engine reads vs. what it reports
The Pico Neural FAQ separates two different observations: what an AI engine reads on a page and what it reports in an answer. Here is how to fill in the two lines and keep them apart.
Open the Pico Neural FAQ, choose your site's page and fill in the two lines, "Read" and "Reported": that is the most direct way to answer questions about ChatGPT and Google without confusing what an AI engine reads on a site with what it reports in its answers. The observed reading concerns the page; the answer is what the engine returns to a question. They are two distinct observations, and the FAQ exists to keep them apart. In the same card you can compare several engines on the same page, because Google's reading of a page and that of the AI engines observed do not coincide with what appears in the answers. "Google AI" is the name in the public list; "Google AI Overviews" is the surface observed, and Gemini and Google AI Overviews are not the same observation.
How to fill in the FAQ card, step by step
The method we propose is an operational proposal from Pico Neural, not a universal standard. Its purpose is to make the comparison between reading and answer verifiable.
- Choose a page and a question. Open the FAQ, select the site's public page and write the exact question you asked the engine. Verifiable criterion: the question must be the same for every engine you compare, otherwise the comparison does not hold.
- Note the "Read" line. Record what the engine reads on the page, with the portion of text or the section it refers to. Verifiable criterion: every entry must be traceable in the observed page; if it is not, flag it as "not traceable" instead of correcting it from memory.
- Note the "Reported" line. Copy the engine's answer and indicate whether it cites a portion present in the observed reading or a portion that is absent. Verifiable criterion: the citation must be compared word for word with the "Read" line.
- Work out the "Observed difference". Describe in one sentence what changes between reading and answer: omission, summary, citation of a portion that is not present, or an answer that does not use the page. Verifiable criterion: the difference must be described without evaluative adjectives.
- Duplicate the card for each engine. If you compare two engines on the same page, fill in two separate cards with the same question. Verifiable criterion: the two cards must show the same text in the "Read" line only if the observed reading coincides; otherwise they must be kept distinct.
An illustrative example with two engines on the same page
This is an illustrative example, not a real case. Imagine a page with three criteria for choosing a plan. In the first engine's card, the "Read" line lists the three criteria; the "Reported" line contains an answer that cites two of them and omits the third; the "Observed difference" is "omission of the third criterion". In the second engine's card, the "Read" line lists the same three criteria; the "Reported" line cites a portion that does not appear in the observed reading; the "Observed difference" is "citation not traceable". The two cards stay separate because the answer is different even though the page is the same.
The point the FAQ makes visible is that reading a page and answering a question are two different events, and they must be noted on two separate lines. When the answer cites a portion you cannot trace in the observed reading, do not rewrite the "Read" line to make it match: add a note in the "Reported" line marked "portion not traceable" and leave the reading as it is. That way the card keeps two pieces of information: what the engine read and what it reported. Duplicating the card for each engine keeps one engine's answer from being attributed to another. If you compare ChatGPT, Perplexity, Gemini, Grok and Google AI, remember that Claude and Grok are measured only after you connect them, and that Gemini and Google AI Overviews are not the same observation. The FAQ does not say which engine is better: it only shows where reading and answer diverge, so you can interpret the card without expectations the facts do not support.
Verifiable limit
The Pico Neural FAQ does not establish which engine reads a page better, nor does it guarantee that an answer will cite the observed reading: it shows the two lines and the difference, and leaves the verification to you. For the full picture on how to use it, see the dedicated page.
Pico Neural FAQ: why ChatGPT and Google read the same page differently