Guide, with original data
How to check every week whether ChatGPT, Gemini and Perplexity recommend your business.
· Measured 1 to 25 September 2026 and 4 October 2026
Put the same buyer questions to ChatGPT, Gemini and Perplexity on the same day each week. Record whether each answer names you, who it names instead and which sites it cites, and judge the trend over several weeks rather than any single answer. A spreadsheet is enough to start; a tracker does the asking and the record-keeping for you.
The repetition is the point, because answers move. In our own data, asked the same question at the next run, the engines named 54% of the same tools again, and 5% of answers came back with an identical list. The sites they cited moved more: 39% came back.
Key takeaways
- Same questions, same engines, same day each week. Record who was named and which sites were cited, not only whether you were.
- Expect movement. Asked again at the next run, the engines named 54% of the same tools; Perplexity was steadiest (67%), Gemini least (45%).
- Sources move more: 39% of cited sites came back, against 67% of the sites on Google's first page for the same questions.
- Check each engine. In our July 2026 study the four assistants read a website in common on only 3 of 50 questions.
- Judge a change by a trend over several weeks, on more than one question.
Why one check is not enough
We ask the same 20 buyer questions on every run. Comparing each answer with the answer to the same question on the same engine at the previous run, from 1 to 25 September 2026:
| Engine | Tools named again | Identical list | Sites cited again |
|---|---|---|---|
| Perplexity | 67% | 10% | 55% |
| Google AI Overviews | 53% | 11% | 35% |
| ChatGPT | 53% | 2% | 27% |
| Claude | 51% | 4% | 47% |
| Gemini | 45% | 0% | 20% |
Tools named again: of the tools an answer named, the share named again at the next run. Identical list: how often the whole list came back unchanged. Sites cited again: the same measure for the sources. On every engine, one reading tells you who was named that day.
Google's first page moved too, but less than the AI answers to the same questions. That is the case for checking the answers themselves every week, rather than inferring them from a rank report.
The engines also disagree with each other. In our AI citation study of July 2026, which put 50 buyer questions in five industries to ChatGPT, Claude, Gemini and Perplexity, all four read at least one website in common on only 3 of the 50 questions. Each industry report publishes its data as a CSV under CC BY 4.0. Being named by one engine tells you little about the others, so the weekly check covers each engine separately.
The weekly check, step by step
- Choose the questions. Write what a buyer types when choosing: “best [what you sell] for [who you sell to]”, comparisons, the problem you solve, your town if you are local. Leave your own name out, because a question that names you is answered with you. Ten questions is a workable start.
- Fix the conditions. Same wording, same engines, same day of the week, with web search on. Use a fresh or logged-out session where you can, and ask from the country you sell in, because answers differ by place.
- Record every answer. One row per question, per engine, per week. Save the answer itself, not only a yes or no.
- Score it per engine. Mention rate is the answers that named you divided by the answers given. Leave out an answer that never came, such as a search where Google showed no AI Overview, rather than counting it as a miss.
- Read the trend, not the week. With about half of the named tools changing from one run to the next, a single drop is often movement, not a verdict. Act on a change that holds for several weeks and shows on more than one question.
- Act on the sources. The sites cited beside the competitors who are named are where to get listed or quoted. Which websites ChatGPT cites shows what that list looks like, engine by engine.
A layout that works for step 3, one row per answer:
| Date | Engine | Question | Named you? | Named instead | Sites cited |
|---|---|---|---|---|---|
| Monday's date | ChatGPT | Best [what you sell] for [who you sell to]? | Yes or no, and where in the answer | The competitors it named | Every site it linked |
The same question, asked on 4 October 2026
We track “How can I check every week whether ChatGPT, Gemini and Perplexity recommend my business?” ourselves. On 4 October 2026, four of the five engines answered it:
| Engine | Tools named, from our list | Sites cited |
|---|---|---|
| ChatGPT | none of the tools we count | openai.com |
| Claude | Semrush, Profound, Otterly.AI, Peec AI, Ahrefs and SE Ranking | rivo.io, seranking.com, position.digital, integrate.io, digitalapplied.com, sonary.com, dageno.ai, therankmasters.com, visiblie.com |
| Gemini | Semrush, Profound, Ahrefs, SE Ranking, LLM Pulse and Mentionova | demandsage.com, ellevate.fr, frizerly.com, keyword.com, seranking.com, sonary.com, sotaventomedios.com, rivo.io, llmpulse.ai, rankability.com, cheers.tech, getintel.ai, inquirer.net, vgraple.com |
| Perplexity | Semrush, Otterly.AI and SE Ranking | usatoday.com, llmpulse.ai, yoast.com, blog.hubspot.com, improvado.io, digitalapplied.com, manilatimes.net, trustmary.com, menafn.com, integrate.io, otterly.ai, airops.com, semrush.com, frase.io |
| Google AI Overviews | No answer: results page not read that day | |
SE Ranking and Semrush were each named by 3 of the 4 engines that answered. ChatGPT named none of the 42 tools we count. Asking the question once, as we did here, is exactly the single reading the rest of this page warns about; the weekly check is what turns it into a line.
Free and paid ways to run it
Free, from the search engines. Two reports show your own pages in their AI features. Neither asks the questions for you, but both are worth setting up before anything else:
- Google Search Console: Generative AI performance report. How often links to your site were shown in Google's generative AI features, as impressions, with filters for text and multimodal searches. Covers AI Overviews and AI Mode.
- Bing Webmaster Tools: AI Performance report. Total citations of your pages in AI answers, the grounding queries the AI used to find them, and citation counts per URL. Covers Microsoft Copilot, AI summaries in Bing and select partner integrations.
Free, one question at a time. Several vendors run a free one-off check; run one on the same day each week and you have a manual tracker. The free checkers, and each one's limits.
Paid, on a schedule. A tracker asks your questions for you and keeps every answer. These state their schedule on their own sites, in their words, checked on 25 September 2026:
- Profound: Runs every tracked prompt daily and captures answers from the browser, as users see them.
- Peec AI: Each client is its own project in one account, with weekly automated reports, dashboards in your own branding, and Looker Studio dashboards clients open through read-only links, no Peec login.
- Surfer: AI Tracker, a Surfer feature that tracks brand mentions, citations and share of voice across five AI engines, refreshed daily.
- GrackerAI: A multi-brand workspace with client-level access controls and weekly white-label PDF reports.
- LLMrefs: Geo-targeting in more than 50 countries and 20 languages, with weekly reports.
- LLM Pulse: A 14-day free trial on its weekly Starter, Growth and Scale plans, plus free tools.
- Wellows: An AI visibility and content platform for brands and agencies that tracks citations daily and drafts pages meant to earn them.
- Omnia: Tracks every prompt every 24 hours.
Daily or weekly matters less than keeping the questions and conditions fixed, so that a change in the line is a change in the answers. Which trackers the engines themselves recommend is in our ranking, linked from the method below.
How we measured this
Our methodology: these are buyer questions we track for our own product in our own workspace, asked as a buyer in the United States. Each figure on this page names the sample it comes from:
- September: 391 answers to 20 buyer questions from five engines, in four runs on 1, 2, 10 and 25 September 2026. They are the same answers our ranking of AI visibility tools counts, asked through the same models.
- 4 October 2026: one run of all 25 questions we track now, including five added on 25 September. 125 questions went out and 104 answers came back. ChatGPT answered 21 of 25, and Google's results page could be read for 9 of the 25 questions, 8 of which showed an AI Overview. Missing answers are left out, never counted as misses. The engines had moved to newer models on 29 September: ChatGPT gpt-5.6-luna, Claude claude-sonnet-5-5, Gemini gemini-3.6-flash and Perplexity perplexity/sonar.
The questions quoted on this page, exactly as the engines were asked them:
- “How can I check every week whether ChatGPT, Gemini and Perplexity recommend my business?”
Cited means the website was among the sources the engine returned for that answer, counted once per answer however many of its pages were used.
Named means the tool appears in the answer's own words, matched the way our ranking matches it. Links are removed first, because a link to a vendor's blog is a source and not a recommendation. A tool named in the question is not counted for it. We match 42 AI visibility vendors plus our own product, so a tool outside that list is not counted however often it appears.
Google's first page is the organic results on page one of Google for the same question typed in full, read in the same visit as the AI Overview. It was captured for 72 of the 80 times a question was asked in September.
Came back compares an answer with the answer to the same question on the same engine at the next run: of the sites, or tools, the first one had, the share the next one had again.
Where this can mislead. We chose the questions, and we sell a tool in this category, so read these as one company's measurement rather than a census. Answers came through each engine's API or Google's results page, which is close to what the consumer apps show but not identical, from one country in English.
Nobody paid to appear on this page, and no vendor saw it before it was published. Our editorial rule is to describe every vendor only in the words of its own site. If a figure or a description here looks wrong, tell us through the contact page.
Where Greater Than Services fits
Our own product, so read this as the vendor's description, not part of the measurement above.
Greater Than Services is a workspace per brand that puts your buyer questions to five AI engines every week, records which answers named you and what they cited instead, shows your Google rank beside each answer, and drafts the fixes. In the count above it was named in none of the 391 answers.
- If you need the weekly check done for you, it puts your tracked questions to the engines once a week and records who each answer named and which sites it cited.
- If you need Gemini and Perplexity as well as ChatGPT, Growth and Agency ask all five: ChatGPT, Google AI Overviews, Claude, Perplexity and Gemini. Starter asks ChatGPT and Google AI Overviews.
- If you need to see what moved, every answer is kept in full, and an email each week tells you what changed.
- If you need to check straight after a fix, on-demand refreshes let you re-ask without waiting a week: two a month on Starter, up to 30 on Agency.
Questions are asked once a week, in English, from the market you choose.
Plans start at $49 a month, billed yearly, for one brand on ChatGPT and Google AI Overviews; Growth and Agency cover all five engines. The three-day trial is on one brand and needs a card.
Questions people ask
How often should I check whether ChatGPT recommends my business?
Every week is a sensible rhythm: often enough to see a change land, and within a month or two enough readings to tell a trend from the engines' own movement. In our data, asked the same question at the next run, the engines named 54% of the same tools again, so judge the line over several weeks rather than one reading.
Why did ChatGPT stop recommending my business?
It may be nothing you did. Answers move between runs even when nobody changes anything: in our data 54% of the tools named came back at the next run, and 39% of the sites cited. Models are updated too. Check whether the sites that cited you are still being cited before changing your pages.
Can I check for free?
Yes, by hand: the same questions, on the same day each week, in a spreadsheet. Google Search Console and Bing Webmaster Tools also report your own pages' appearances in their AI features for free.
If ChatGPT recommends me, do I need to check Gemini and Perplexity?
Yes. The engines read different sources: in our July 2026 study of 50 buyer questions, ChatGPT, Claude, Gemini and Perplexity read at least one website in common on only 3 of them. Being named by one says little about the others.
Should I put my business name in the questions?
No. A question that names you is answered with you, so it says nothing about whether you are recommended. Ask what a buyer asks before they know your name.
Can Greater Than Services run the weekly check for me?
Yes. It is our own product: once a week it puts your tracked questions to the engines your plan covers and records who each answer named and which sites it cited. Starter covers ChatGPT and Google AI Overviews; Growth and Agency cover all five.