To check your AI-search shortlist position, ask ChatGPT, Perplexity and Gemini the same five buyer prompts. For every answer, record which vendors get named and which pages get cited. Then count how often you appear and pick the page that most often decides the answer. Google rankings do not predict the result, so you have to ask.
Key takeaways
- Five buyer prompts across three engines give 15 answers. That is enough to see who owns your category's shortlist.
- Record two things per answer: the vendors named and the sources cited. The sources tell you what to fix.
- Share of answer is the number of answers that name you, divided by the number of answers you ran.
- Fix one page first: the source type that appears most often and that you can credibly publish or earn.
- Blesso's own scan scored 0 across three engines on seven tracked prompts. Starting at zero is a position you can measure and change, and no result is promised.
Why Google rank does not predict the AI shortlist
A Google ranking tells you where your page sits for one query. An AI answer is a different product. The assistant writes a short list of vendors, and it draws on pages it retrieves for that prompt, which are often roundups, guides and directories rather than vendor home pages. You can rank well for your own category keyword and still be absent from that list, because the list is built from other people's pages.
Blesso is a done-for-you B2B go-to-market service with an AEO / GEO line that tracks buyer prompts weekly. It also has this problem itself. Blesso's workspace scan on 26 October 2026 gave blesso.ai a visibility score of 0 across ChatGPT, Perplexity and Gemini on seven tracked prompts. Blesso was named in none of those answers. That is the starting point this method is built for: a measured zero, followed by a plan.
The check in five steps
- Write five buyer prompts. Use the questions a buyer types before they have a shortlist.
- Run each prompt on three engines. ChatGPT, Perplexity and Gemini, in a fresh chat each time.
- Record the vendors named. One row per answer.
- Record the sources cited. Note the page type and the publisher.
- Count your share and choose the first page to fix.
1. Write five buyer prompts
Cover the stages a B2B SaaS buyer moves through. A spread of five works well:
- A category question: "What is the best [category] for a [size] [industry] company?"
- An alternatives question: "What are the alternatives to [the leading vendor]?"
- A price question: "How much does [category] cost per month?"
- A fit question: "Which [category] works for a team with no [role]?"
- A comparison question: "[Your category option] versus [substitute]: which should I choose?"
Write them in your buyer's words, not your own product language. Never put your company name in a prompt. A prompt that names you only tests whether the engine has heard of you, which is not the shortlist question.
2. Run each prompt on three engines
Open a new chat for every run so earlier answers do not shape the next one. Run all five prompts on ChatGPT, Perplexity and Gemini. That gives 15 answers. Do it in one sitting and note the date, because engines change their answers over time.
3. Record the vendors named
Keep a sheet with one row per answer. Copy the vendor names in the order the engine lists them. Order matters less than presence, but note it when an engine marks one vendor as the top pick.
| Prompt | Engine | Vendors named | Are you named? |
|---|---|---|---|
| Alternatives to [leader] | ChatGPT | Vendor A, Vendor B, Vendor C | No |
| Alternatives to [leader] | Perplexity | Vendor A, Vendor D | No |
| Alternatives to [leader] | Gemini | Vendor B, Vendor C, Vendor E | No |
The rows above are a blank format with placeholder names, not results. Fill yours with what you see. Blesso's scan shows why the full list matters: on an "alternatives to AEO services" prompt, the engines returned between 10 and 18 named providers, and Blesso appeared in none. A long list means many seats, and also many competitors for each one.
4. Record the sources cited
Perplexity shows its sources beside the answer. ChatGPT and Gemini show them in many answers, and you can ask directly: "Which pages did you use for each vendor?" Add a second sheet with one row per cited page: the address, the publisher, and the page type.
| Page type | What to look for | Who controls it |
|---|---|---|
| Roundup or "best of" list | Titles with "best", "top" or a year | A third party, often a vendor in the list |
| Alternatives or versus page | "[Vendor] alternatives", "[A] vs [B]" | A vendor or a publisher |
| Guide or pricing explainer | "How to choose", "What it costs" | Anyone, including you |
| Directory or review site | A listing page with ratings | The platform, with your profile input |
| Vendor's own page | A product or service page | The vendor |
| Community thread | Reddit or forum answers | The community |
A cited page is a lead, not proof. An engine can list a source without quoting it, so read the page and confirm that your target vendors appear on it before you act.
Count your share of answer
Share of answer is the number of answers that name you, divided by the number of answers you ran. Blesso's AEO / GEO line monitors this weekly across engines. You can compute a first version by hand.
Here is a made-up example, labeled as such. You run 5 prompts on 3 engines, which gives 15 answers. Your company is named in 3 of them. Your share of answer is 3 divided by 15, which is 20%. If one rival is named in 9 of the 15, their share is 60%. The two numbers, taken side by side, tell you more than either alone.
Compute it three ways: overall, per engine, and per prompt. A zero on one engine with a hit on another points to a source that one engine reads and the other does not. A zero on the price prompt only points to a missing pricing page.
Choose which page to fix first
Do not try to fix every gap. Pick one page using three questions about your source sheet.
- Which page type appears most often? If roundups dominate, your path is to be included in those roundups or publish your own honest comparison. If vendor pages dominate, fix your own page.
- Which of those can you change? You control your own pages. You only influence third-party pages, and that takes outreach and time.
- Which prompt is closest to a buyer's decision? A missing seat on "alternatives to [leader]" costs more than a missing seat on a definition question.
In most cases the first fix is a page you own that answers one of your five prompts directly, in the first two sentences, with a clear statement of what you do and for whom. Then re-run the same five prompts and compare. If the cited pages are all third-party roundups and you own none of them, put "get included in the top two roundups" next to that page on your list, and treat it as slower work.
Blesso's own scan points the same way. In the workspace data, rivals win on listicles and guides, not on vendor landing pages. That is a finding about one category, so check yours against your own sheet.
What to do after the first fix
Repeat the same prompts weekly, with the same wording, so the numbers compare. Log the date, the engine and the share. Expect movement to be slow and uneven. No tool or service can promise that a given page will be cited, and Blesso does not promise a result. What a weekly log gives you is evidence of what changed after each edit.
Blesso runs this tracking as part of its AEO / GEO service: buyer prompt mapping, weekly share-of-answer monitoring, and answer-ready pages. Engagements start from $2k a month with a three-month minimum. Do the manual check first, because it shows whether you need help.
FAQ
How many prompts do I need to check my AI visibility?
Five prompts on three engines, or 15 answers, is enough for a first read. It shows who owns your category's shortlist and which sources feed it. Add more prompts later, one buyer stage at a time.
Why does ChatGPT name competitors that rank below me on Google?
Because ChatGPT builds its list from the pages it retrieves for that prompt, not from your search rank. Those pages are often roundups and guides that mention a competitor and not you.
What is share of answer?
Share of answer is the number of AI answers that name your company, divided by the number of answers you ran. If you run 15 answers and appear in 3, your share is 20%. Track it with the same prompts every week.
Should I put my company name in the test prompts?
No. A prompt that names you tests brand recognition, not shortlist position. Use the questions a buyer asks before they know any vendor, so the answer shows who gets named without being asked.
How often should I re-run the check?
Weekly is a good rhythm for a category that moves. Use identical prompt wording and the same engines each time. Answers vary between runs, so look at the trend over several weeks, not one result.
Want the weekly number without the spreadsheet?
Bring your five prompts to a call. Blesso will show how it maps buyer prompts and tracks share of answer for your category.