I checked 15 AI search guides

I searched Google for three things people ask about AI search: "llm seo", "ai visibility" and "how to rank in chatgpt". For each search I took the first five guides.

We were writing guides on the same three topics for Shruwd (https://shruwd.io/), the tool we are building, so I checked the 15 on a few points our own guides would cover.

Four checks came out of it that you can run on any AI search guide before you act on it.

1. Is there a source behind the number?

Each of the five "llm seo" guides recommends schema markup at least once. Two also say schema is on close to every page ChatGPT cites.
Neither links a source.

A matched study has tested if adding schema brings more citations to pages that already get some. Ahrefs followed 1,885 such pages as their owners added schema from August 2025 to March 2026, and compared them with 4,000 control pages. The effect on citations in ChatGPT and Google's AI Mode, in Ahrefs' words: "Statistically indistinguishable from zero."

Even if the two guides are right, that alone wouldn't show if schema actually helps. Sites that look after their SEO usually add it, so it would turn up on cited pages either way.

So I wouldn't put schema at the centre of an AI search plan on the strength of an unsourced number.

2. Does the number come with a range?

Three of the five "ai visibility" guides recommend tracking a single visibility score, and none of the five shows a range for any number, or says how big a change has to be before you trust it.

At the end of 2025, 600 volunteers ran 12 prompts through ChatGPT, Claude and Google's AI for a SparkToro and Gumshoe study, 2,961 runs in total. For ChatGPT and Google's AI, the authors put the chance of two responses naming the same list of brands at under 1 in 100.

A number from one check can move a lot while nothing about your site changes. If you're named in 10 of 30 answers, the 95% range for your real rate is roughly 19% to 51%.

Semrush's head of organic and AI visibility, Sergei Rogulin, writes that a share of voice "that swings between 20% and 40% over a day is normal". His advice is to report it with a margin.

Before you react to a change in any AI visibility number, ask how many answers it came from. With 30, a swing of 10 points can be noise.

3. Does it cite the vendor's own documentation?

One of the five "how to rank in chatgpt" guides links OpenAI's crawler documentation, and none of them tells you what to make of ChatGPT-User in your logs. OpenAi's page describes several crawlers and three are worth telling apart:

OAI-SearchBot gathers pages for ChatGPT's search. Block it and your site "will not be shown in ChatGPT search answers, though can still appear as navigational links."
ChatGPT-User visits a page on behalf of someone using ChatGPT. For that reason, OpenAI says robots.txt rules "may not apply" to it.
GPTBot crawls content OpenAI may use to train its models, and OpenAI's page doesn't tie it to search.

My reading of a ChatGPT-User visit: someone using ChatGPT asked for something that needed that page, probably at that moment. That makes the visit one of the few traces a ChatGPT answer leaves in a server log.

Two things to check on your own site:

A robots.txt rule meant for GPTBot shouldn't also block OAI-SearchBot.
Before you count a ChatGPT-User visit check its IP against the ranges OpenAI publishes. Any request can call itself ChatGPT-User.

4. Is the check dated?

AI answers can shift in weeks. In Semrush's tracking, the share of ChatGPT responses citing Reddit dropped from nearly six in ten in early August 2025 to about one in ten by mid September.
A guide with no date on its checks gives you no way to tell whether they still hold.

Limits of this check

This was one day (18 September 2026), three searches and 15 pages. I saved each page's HTML and searched its text for the points above, rather than reading each page in full. I wouldn't generalise from it.

We wrote our own guides on the same three topics after these checks: