Field notes
How ChatGPT picks AI visibility tools: what it searched for 7 buyer questions
On 9 October 2026 we asked ChatGPT the questions a small business asks when it looks for an AI visibility or GEO tool, one fresh chat per question, and logged what it searched for, which pages it read and why it recommended each tool. Citelane was not recommended in any answer. Here is what we saw.
Updated October 9, 2026 · Citelane
In our test, ChatGPT ran between 2 and 22 web searches per question, read 19 to 182 pages, and ranked tools mostly from the vendors’ own pricing and feature pages. It favoured pages that state plainly which assistants are covered, the real price of covering all of them (add-ons included), prompt limits, how often checks run and whether the tool gives page-specific fixes. Citelane appeared in none of the seven answers.
How we ran the test
- ChatGPT on the web, default model, temporary chats so past conversations would not steer the answers. One new chat per question, 9 Oct 2026.
- After each answer we asked ChatGPT to list the exact web searches it ran, how it split the question and why it picked those tools. ChatGPT does not show raw searches in the interface, so the search lists below are its own account of that turn.
- The page counts come from the search results attached to each conversation (pages ChatGPT retrieved, not only the ones it linked).
One run, one account, one day. AI answers vary between runs, so read this as a set of observations, not a ranking of tools.
The seven questions and who ChatGPT recommended
| Question (as asked) | Searches | Pages retrieved | Recommended |
|---|---|---|---|
| How do I check whether ChatGPT recommends my business when potential customers ask for products like mine? | 2 | 24 | No tools: a manual method (20–30 non-branded prompts, repeated; mention, top-3 and citation rates) |
| What are the best AI visibility tracking tools (GEO tools) for small businesses to monitor brand mentions in ChatGPT, Perplexity, Gemini, and Claude? | 17 | 98 | LLMrefs, Otterly.AI, Rankscale, Promptwatch, Ahrefs Brand Radar, Peec AI |
| Is there a free AI search visibility checker that lets me see if ChatGPT recommends my brand without signing up? | 2 | 19 | Ahrefs AI Visibility Checker, RankSurf, Rank.ai, GeoBuddy |
| What are the best affordable alternatives to Otterly AI, Profound or Peec AI for a small business? | 22 | 182 | AI Peekaboo, Qwairy, Rankscale, Promptwatch, SE Visible; HubSpot AI Search Grader as a free one-off |
| What is the cheapest LLM brand monitoring tool, under $50/month, for ChatGPT, Gemini and Perplexity? | 8 | 81 | Verified Tools, Astiva AI, LLM Visibility, Otterly.AI |
| Why does ChatGPT keep recommending my competitors instead of my brand, and how do I get recommended? | 3 | 33 | No tools: OpenAI’s own documentation only |
| What is the best AI visibility audit tool that gives actionable GEO recommendations for my pages, not just dashboards? | 8 | 81 | OtterlyAI, Scrunch, AthenaHQ, Profound, Peec AI |
What it searched for
The searches were long and specific. For tool questions ChatGPT first searched the category, then searched each candidate’s own site for pricing and engine coverage. Examples, as ChatGPT reported them:
- best AI visibility tracking tools ChatGPT Perplexity Gemini Claude small business pricing 2026
- site:otterly.ai pricing lite Claude Gemini add on
- site:llmrefs.com pricing eight engines 79
- free AI visibility checker ChatGPT brand mentions no signup free tool 2026
- affordable AI search visibility tracking alternatives Otterly Profound Peec AI pricing 2026 small business
- LLM brand monitoring all three ChatGPT Gemini Perplexity under $50 pricing plans
- best AI visibility audit tools page level actionable GEO recommendations 2026
- site:help.openai.com ChatGPT search how websites appear search results publishers OAI-SearchBot
For the two “how do I” questions it searched only OpenAI’s help centre and crawler documentation, and named no third-party tools.
Why it picked those tools
When we asked, ChatGPT described its selection rules. They differed by question:
- Best tools for four assistants: the real price of a plan that covers ChatGPT, Perplexity, Gemini and Claude, not the entry price. It added up per-engine add-ons itself. Then prompt allowance, daily versus weekly checks, and citation tracking.
- Free checker: three hard filters (free, no sign-up, actually checks ChatGPT), then whether you can enter your own buyer question, see competitors and read the real answers rather than a score.
- Affordable alternatives and under $50: an entry price in the $30–$100 range (or under $50 including add-ons and billing period), ongoing monitoring rather than a one-off score, and coverage of several assistants.
- Actionable audits: page-level URL audits and specific edits ranked highest, then crawler and HTML visibility checks, competitor citation gaps, re-auditing after a change, and a small-team price. It contrasted a weak audit (“score 62/100”) with an actionable one (“add a concise answer below the H2”).
It weighted official pricing and product pages above third-party lists, because prices and engine coverage change often, and said several times that its rankings were feature-based, not accuracy tests.
What kind of pages it read
- Vendor pricing pages, often found with a site: search for the price or engine name.
- “Best X” and “X alternatives” lists, most of them published by vendors in the same market, plus independent comparison sites. On the alternatives question alone it retrieved 182 pages from more than 50 domains.
- Free tool pages with titles that repeat the question, such as “Free AI Visibility Checker — See If ChatGPT Recommends Your Brand”.
- Official documentation (help.openai.com, OpenAI’s crawler docs) for how-it-works questions.
What we changed on Citelane after the test
Citelane was missing from every answer and every retrieved list. Our pages described the product, but not in the terms ChatGPT compared on. So we rewrote them to state the facts it looked for:
- Which assistants each tier covers, by model name. The test also showed that "actually checks ChatGPT" is a hard filter for free checkers, so on 9 Oct we added GPT-5 mini with web search to the free no-sign-up check (it already covered Perplexity Sonar and Gemini).
- The price of covering all four assistants (Free plan and Starter $29/month include all four in every full audit) and the volume trade-off.
- Questions per audit (up to three), refresh frequency (weekly, not daily), and what the audit produces for a page: one fact-checked edit and a check that it is live.
We plan to re-ask the same seven questions after the pages are recrawled and compare.
What this suggests for your own product
- Write the facts an assistant compares on where a crawler can read them: what is covered, the full price, limits, frequency and what is not included.
- Use the words buyers use in their questions (“free”, “no sign-up”, “alternative to X”, “under $50”) in titles and headings, but only where they are true.
- Keep a dated pricing and feature page; assistants search vendor sites directly to confirm.
- Find the comparison lists that assistants retrieve for your category and make sure the facts about you there are correct.
- Re-ask the same questions later. One answer is a sample, not a verdict.
Ask live AI assistants your buyers’ questions and see who they recommend. About a minute, no sign-in.
Run the free diagnosisQuestions
Which AI visibility tools does ChatGPT recommend?
In our 9 Oct 2026 test it named LLMrefs, Otterly.AI, Rankscale, Promptwatch, Ahrefs Brand Radar and Peec AI for four-assistant tracking; Ahrefs, RankSurf, Rank.ai and GeoBuddy as free checkers; AI Peekaboo, Qwairy and SE Visible as affordable alternatives; and OtterlyAI and Scrunch for actionable audits. Answers vary between runs.
Can I see what ChatGPT searched for?
Not in the interface. In our test, asking a follow-up in the same chat (“list the exact web search queries you ran”) returned a list of searches, which is ChatGPT’s own account of that turn.
Does ChatGPT trust vendor pages or review sites more?
For prices and features it said it preferred vendors’ official pricing and product pages, and used third-party lists mainly to discover candidates.