An AI recommendation self-test means asking AI assistants the questions your buyers ask and recording whether your business is named. In my October 26, 2026 test, 12 of 42 answers named one of my 7 businesses. Hand-run tests are free but are snapshots: answers shift run to run, and nobody can promise a recommendation.
Bottom line
Pick the self-test on this page if you have about 30 to 60 minutes and want to see for yourself, in your own browser, what the assistants say.
Pick the free 9-answer test from Get Found with AI (my business) if you would rather I run it logged out on ChatGPT, Gemini and Google AI Mode and send you the 3 fixes I would make first.
Skip paid AI visibility trackers unless you have already run a free test, fixed the basics, and need daily tracking across many prompts or locations.
Want to go straight to it? Jump to the 5-prompt self-test. If you would rather hand it off, request the free 9-answer AI recommendation test from Get Found with AI. That is my business, and I run every test myself.
Key facts
- 12 of 42 Gemini and Grok answers named one of my 7 businesses (my test, 2026-10-26)
- Same question, three runs: the top pick changed in 6 of 14 run sets (my test, 2026-10-26)
- 21 of 24 Kansas City answers named or linked a directory or review site (my test, 2026-10-26)
- Robots.txt changes for OAI-SearchBot can take about 24 hours to register (OpenAI)
- Google users clicked a traditional result in 8% of AI-summary visits vs 15% without (Pew Research Center)
What an AI recommendation test is, and what it is not
An AI recommendation test is simple. You ask AI assistants the unbranded questions a buyer would ask, like "best roofer in Kansas City," and write down whether you get named.
It does not track your rank in Google's blue links. A branded "what do you know about me" check on its own does not count either, because buyers rarely ask that. It is also no SEO audit, and it promises nothing.
Keep the three Google and OpenAI surfaces straight. AI Overviews is the summary at the top of a normal Google search. AI Mode is Google's separate chat-style search. ChatGPT search is OpenAI's web search inside ChatGPT. Answers also change with location, saved memory and whether you are signed in, as OpenAI's help page explains.
Take the self-test
Plan on about 30 to 60 minutes to run 25 answers and tick the checklist. Everything runs in your browser, and nothing is sent to me or anyone else.
Run the 5-prompt AI recommendation self-test
Nothing you type leaves this page. Your answers are saved only in this browser.
Unbranded buyer answers make up half the score. The rest covers branded accuracy, crawler access, listings, reviews and answer-ready pages. See how the score is weighted for the full breakdown.
Our data: what Gemini and Grok said about my 7 businesses
On October 26, 2026, I asked one ready-to-buy "money question" for each of my 7 businesses, 3 times each, on Gemini and Grok through their APIs with live web search on. That gives 6 answers per business and 42 in total, scored by script for whether the business was named and who the top pick was. These were API answers, not the consumer apps, and I did not run ChatGPT, Perplexity or Google AI Mode, so you fill those in by hand in the self-test.
| Business (mine) | Money question | Named (of 6) | Top pick most often |
|---|---|---|---|
| Crypto Guidance | who is the best crypto consultant in Kansas City | 6 | Crypto Guidance (6 of 6) |
| KC Roofing Directory | best directory for finding roofers in Kansas City | 6 | KC Roofing Directory (6 of 6) |
| Kansas City SEO Services | best local SEO company in Kansas City | 0 | WiseGuys Digital Marketing (4 of 6) |
| Email Marketing KC | best email marketing agency in Kansas City | 0 | Blue Gurus (3 of 6, all on Grok) |
| Blur AI Solutions | best AI receptionist for small business | 0 | Goodcall (3, Gemini) and Rosie (3, Grok) |
| ContactLint | best email validation API | 0 | ZeroBounce (5 of 6) |
| Get Found with AI | best AI search optimization agency | 0 | First Page Sage (2 of 6, Grok) |
Only 12 of 42 answers named one of my businesses, and both wins came from narrow Kansas City questions. The top pick also moved around: same question, same assistant, 3 runs in a row, and it changed in 6 of 14 run sets (6 / 14 = 43%). For the 4 Kansas City questions, 21 of 24 answers named or linked a directory or review site: Clutch in 13, Thumbtack in 10 and Yelp in 7. All 7 of my sites publish llms.txt and name OAI-SearchBot in robots.txt, yet 5 of the 7 were named 0 times.
Download the raw scores: ai-recommendation-self-test-2026-10-26.csv. This is 7 businesses on one day, so treat it as a pattern to check, not a law.
How ChatGPT, Claude, Perplexity and Google decide who to name
Nobody outside these companies knows the full recipe, and nobody can guarantee that an AI assistant will recommend a business. What the vendors do publish is who their crawlers are and what makes a page eligible.
Crawler access, per assistant. OpenAI runs separate bots. Its bot documentation says:
"OAI-SearchBot is used to surface websites in search results in ChatGPT's search features. Sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers, though can still appear as navigational links."
OpenAI, Overview of OpenAI Crawlers
GPTBot is for training and is a separate setting. A robots.txt change can take about 24 hours to register. Anthropic lists 3 robots: ClaudeBot for training, Claude-User for fetching pages when a person asks, and Claude-SearchBot for search quality. Blocking Claude-User may cut your visibility in user-directed search. Perplexity says PerplexityBot surfaces and links sites in its results and is not used to train foundation models.
Being indexed. Google says there are no additional requirements to appear in AI Overviews or AI Mode. To show up as a supporting link, your page has to be indexed and eligible for a snippet in normal Google Search. For local results, Google says ranking is mainly based on relevance, distance and popularity.
Location in the question. ChatGPT may guess your area from your IP address. If the local results are for the wrong area, OpenAI suggests putting your city, neighborhood or postal code in the question. That is why every prompt in the self-test includes your area.
Outside proof. In my Kansas City answers, the assistants leaned on directories and review sites in 21 of 24 answers. If Clutch, Thumbtack or Yelp is where the answer gets its list, that is where you need to be.
Why this matters. Pew Research Center tracked 900 U.S. adults in March 2025. An AI summary showed up on 18% of their Google searches. With a summary, people clicked a traditional result in 8% of visits, versus 15% without one. They clicked a link inside the summary in just 1% of visits. If the summary names someone else, fewer people scroll down to find you.
How to run the test, step by step
1. Write your money question
Write the question a buyer types when they are ready to pay, in their words, with your area in it. "Emergency plumber in Overland Park" works. "Plumbing solutions provider" does not.
Time: about 10 minutes.
Done when: you have one service phrase and one area the tool can turn into 5 prompts.
2. Open a private window
Sign out, or open a private or incognito window, so memory and account history do not tilt the answer toward you.
Time: about 1 minute.
Done when: each assistant is open with no account signed in, where it allows that.
3. Run the 4 unbranded prompts in 5 assistants
Paste each prompt into ChatGPT, Claude, Perplexity, Gemini and Google AI Overviews or AI Mode. Mark each answer Picked (recommended you), Named or Missing. If you skip an assistant, mark it Not run so it stays out of the math.
Time: about 20 to 35 minutes for 20 answers.
Done when: all 20 unbranded cells are marked.
4. Run the branded check last
Ask each assistant what it knows about your business by name. Do this last so it cannot shape the earlier answers in the same session. Mark any wrong hours, services or area as "Wrong."
Time: about 5 to 10 minutes.
Done when: all 5 branded cells are marked.
5. Record the sources each answer cites
Write down every directory, review site or article the answers link to. That list is your to-do list.
Time: about 5 to 10 minutes, done alongside steps 3 and 4.
Done when: you have a short list of the sites that keep showing up.
6. Fix, then re-test on a set date
Work on the 3 fixes the tool flags. Then put a date on your calendar each month to run the same prompts again, so you compare like with like.
Time: about 30 to 60 minutes per monthly re-test.
Done when: the next test date is on your calendar and you have saved your copied results.
Mistakes that cost the most
Blocking OAI-SearchBot
If your robots.txt, host or CDN blocks OAI-SearchBot, OpenAI says your site will not be shown in ChatGPT search answers. That is 100% of those answers, apart from navigational links. Even after you fix it, OpenAI says it can take about 24 hours to register.
Fix: allow OAI-SearchBot in robots.txt, make sure your host allows OpenAI's published IP addresses, and wait a day before you re-test.
Trusting one screenshot
One answer tells you very little. In my test, the top pick changed in 6 of 14 run sets: 6 / 14 = 43%. A single screenshot, good or bad, could easily be the odd one out.
Fix: run each prompt in several assistants, and look at the pattern across all 25 answers.
Only testing your own name, or paying for a tracker first
Asking "what do you know about my business" tells you about accuracy, not whether buyers will hear about you. Paying for tracking before you know your baseline costs real money: High Street's guide says the tracker it recommends starts at $49 a month, so $49 x 12 = $588 a year, before you have run the free 25-answer self-test.
Fix: run the unbranded prompts first, free, and only pay for tracking once you know what you are tracking.
DIY self-test vs free checker vs paid tracker vs done-for-you test
| Option | Cost | Time | Your effort | Best for |
|---|---|---|---|---|
| This self-test (on this page) | Free | About 30 to 60 minutes | High: you run 25 answers and tick 12 checklist items | Owners who want to see the answers first-hand |
| BrightLocal Local AI Visibility Checker | Free | Short | Low: enter your business details | A quick look at how you appear across ChatGPT and Google's AIs |
| Paid daily trackers | High Street says Cituna starts at $49 a month | Ongoing | Low once set up | Brands tracking many prompts or locations over time |
| Get Found with AI free 9-answer test (my business) | Free | I run it and email the results | Low: a six-field form | Owners who want 9 logged-out answers scored, plus 3 first fixes |
| Hiring an AI search agency | Varies, ask for a written quote | Varies | Low to medium | Businesses ready to pay for ongoing work after a baseline test |
My pick for most Kansas City service businesses: run the self-test on this page first. It is free, it shows you the actual answers, and it tells you which 3 fixes carry the most weight. If you get stuck or run out of time, request the free 9-answer test from my business, Get Found with AI. If you are weighing paid help, my page on what SEO costs in Kansas City covers budgets.
How the score is weighted
| Part | Weight | What it measures |
|---|---|---|
| Unbranded answers | 50 | 4 prompts x 5 assistants = 20 answers, 2 points max each |
| Branded accuracy | 10 | Whether 5 assistants get your details right when asked by name |
| Crawler access and indexing | 10 | AI search bots allowed, pages indexed in Google |
| Listings and name, address, phone consistency | 15 | Google Business Profile, directories, matching details everywhere |
| Reviews | 10 | Reviews on the sites answers cite, and replies to them |
| Answer-ready pages | 5 | Schema, an answer-first service page, llms.txt |
Unbranded answers carry half the weight (50 of 100) because that is what a buyer actually sees. In my data, clean crawler files did not lead to mentions on their own: 5 of 7 sites with llms.txt were named 0 times.
Disclosure: Get Found with AI is my business, and it is one of the options compared above. The self-test on this page is free and sends nothing to me. The data table covers my own businesses only, not clients.
What your score means
0 to 39: AI assistants rarely name you
What it means: When buyers ask, you are mostly missing from the answers, or the basics like crawler access and listings are not in place.
Do next: Check robots.txt and indexing first, then claim and match your listings on the directories the answers cited.
Trade-off: Fixing basics is cheap but slow to show up, and no one can say when an assistant will start naming you.
40 to 69: Named sometimes, not reliably
What it means: Some assistants name you on some prompts. Your foundation exists, but outside proof is thin.
Do next: Work on the assistants that scored lowest, add reviews on the cited sites, and write one answer-first page for your money question.
Trade-off: Results vary run to run, so a mid score can move up or down with no change on your end.
70 to 100: Named often, keep re-testing monthly
What it means: You show up across most assistants and prompts, and your details are mostly right.
Do next: Re-test monthly on a set date, keep reviews coming, and fix any wrong details right away.
Trade-off: A high score is a snapshot. In my test, top picks changed in 6 of 14 run sets.
- If you are a roofer or contractor: start with Thumbtack, Angi and BBB listings. Watch out for wrong service areas in the branded check.
- If you are a restaurant or shop: start with Google Business Profile and Yelp, and reply to reviews as Google recommends. Watch out for old hours showing up in answers.
- If you are a professional service like a CPA or lawyer: start with the directories and review sites your field uses, plus a clear answer-first page. Watch out for assistants mixing you up with a similar firm name. The AI readiness quiz is a good next step.
- If you are a multi-location brand: start by running the self-test once per city with that city in the prompt. Watch out for one location's details bleeding into another's.
Who this is not for
National ecommerce brands with thousands of products. A 5-prompt test cannot cover a big catalog. You are better served by a paid tracker built for many prompts at once.
Businesses that need paid leads this week. AI visibility work moves slowly and nothing is promised. Run paid ads now, and start here in parallel. If missed calls are the real leak, the missed call revenue calculator shows what they cost you.
Regulated advice firms that cannot change listings without compliance review. You can still run the test, but the fixes need sign-off. Take the copied results to your compliance team before changing anything.
How I put this together
On October 26, 2026, I read the pages that come up for "does chatgpt recommend my business" and "how to check if chatgpt recommends your business."
IMPACT gives 5 practical ways to get recommended and has strong content advice. It has no self-test, no scoring and no local data. Zarla has good local framing and tells you to re-check on a schedule, but the check is one short instruction with no grader. FactoryJet lists 7 signals with a ranked playbook and a timeline. It does not cover run-to-run variance or link the vendors' crawler docs. High Street has the strongest how-to-check method, with a 2/1/0 grid and the branded check saved for last. It is written for UAE businesses and points to a paid tracker.
This page adds: a grader that scores 25 answers and 12 checklist items in your browser, my own logged data from 42 answers with the variance shown, direct links to the OpenAI, Google, Anthropic and Perplexity docs, and Kansas City questions. More about me and how I build is on my about page. For other agent work, see AI agents in Kansas City, and if you are choosing help, read how to pick an SEO consultant in Kansas City.
Frequently asked questions
Does ChatGPT recommend local businesses in Kansas City?
Yes, it names local businesses when you ask with a city or neighborhood, and OpenAI says it may use an approximate location from your IP address. In my October 26, 2026 test on Gemini and Grok, 21 of 24 Kansas City answers also named or linked a directory or review site. Which business gets named can change from run to run.
How do I check if ChatGPT recommends my business?
Open a private window, ask ChatGPT the unbranded question a buyer would ask with your city in it, and note whether you are named. Repeat it in other assistants, then ask about your business by name last. The self-test on this page writes the 5 prompts and scores the 25 answers for you.
Why does ChatGPT recommend my competitor and not me?
Usually because the sources the answer leans on mention them and not you. In my Kansas City answers, Clutch showed up in 13 of 24, Thumbtack in 10 of 24 and Yelp in 7 of 24. Check which sources the answer cites, then work on being listed and reviewed there.
Is there a free AI visibility checker?
Yes. The self-test on this page is free and runs in your browser, BrightLocal offers a free Local AI Visibility Checker, and my business Get Found with AI runs a free 9-answer AI recommendation test. None of them can promise placement.
How often should I re-test?
Monthly is a sensible default, on a set date, with the same prompts. Answers shift between runs, so compare patterns, not one screenshot. After a robots.txt fix, wait at least a day, since OpenAI says changes take about 24 hours to register.
Does an llms.txt file get my business into ChatGPT?
Not on its own. All 7 of my business sites publish an llms.txt file and name OAI-SearchBot in robots.txt, yet 5 of the 7 were named 0 times in my October 26, 2026 test. Crawler access is needed, but outside proof like directories and reviews counts too.
Can anyone guarantee ChatGPT will recommend my business?
No. OpenAI's own help page says placement in ChatGPT search is not promised, and in my test the top pick changed in 6 of 14 run sets. Good work improves the chance of being named or cited, but nobody controls the answer.
Start with the self-test and copy your results. Then request the free 9-answer AI recommendation test from Get Found with AI, my business, or email your score to [email protected] and I will tell you which fix I would make first. More free tools are on the tools page.
Sources
- OpenAI: Overview of OpenAI crawlers (OAI-SearchBot, GPTBot, ChatGPT-User)
- OpenAI Help Center: Searching the web with ChatGPT
- Google Search Central: AI features and your website
- Google Business Profile Help: Tips to improve your local ranking on Google
- Anthropic: Does Anthropic crawl data from the web, and how can site owners block the crawler?
- Perplexity: Perplexity crawlers
- Pew Research Center: Google users are less likely to click on links when an AI summary appears (July 22, 2025)
- High Street: How to check if ChatGPT recommends your business (updated October 25, 2026)
- BrightLocal: Free Local AI Visibility Checker
- IMPACT: How to get your business found and recommended on ChatGPT
- Zarla: How to get your business recommended by ChatGPT (August 25, 2026)
- FactoryJet: How to get ChatGPT to recommend your small business (June 7, 2026)
- Get Found with AI (my business): free 9-answer AI recommendation test
- My AI recommendation test of 7 of my businesses on Gemini and Grok, October 26, 2026 (CSV, 42 answers)
What changed in this guide
- 2026-10-26: First published. 5-prompt, 5-assistant self-test and grader (scored out of 100), my October 26, 2026 test of 42 Gemini and Grok answers with CSV, crawler rules from OpenAI, Google, Anthropic and Perplexity, and a clean Markdown copy for AI agents.