
What matters first.
- An AI recommendation means an engine makes you its pick in its answer to one buyer question. You win that slot from a named rival, question by question and engine by engine, not from “AI” in general.
- Start with the questions where ChatGPT, Perplexity, Gemini, Claude or Google AI Overviews already recommend a rival. Read the reason and the source behind it in the stored answer, and close first the one gap that loses on the most engines.
- Count the switch only when FameScan’s verdict on your recommendation rate reads Real change, not Within noise or Too few answers. The verdict tests the selected period against the one before it, and each rate shows its 95% interval.
How to get AI to recommend your business: the pages ChatGPT, Perplexity, Gemini, Claude or Google AI Overviews read for your buyer’s question must name you and give a reason to choose you. Start where they make one named rival their pick today: find those questions, read the reason in the answer, and close that one gap first. Count the switch only when your recommendation rate moves beyond normal variation, not when one answer flips.
FameScan measures how your brand and your named rivals appear in AI answers to the questions your buyers ask, across ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews: it stores every answer with its sources, separates mentions, citations and recommendations, and ranks the next move in its Action Center. Facts about the engines come from their official pages, opened on November 14, 2026.
How to get AI to recommend your business?
To get ChatGPT or any other AI engine to recommend your business, the pages it reads must name you as the answer to your buyer’s exact question, and give a reason. Start with the questions where it recommends someone else today. An AI recommendation is the slot an engine gives one or a few brands in its answer to one buyer question; you win it from a named rival, question by question and engine by engine, not from “AI” in general.
Most guides on how to get ChatGPT to recommend your business answer with the same checklist: collect reviews, keep your business details consistent, fill in your Google Business Profile, add structured data, get mentioned on strong sites. That helps. But the checklist treats a recommendation as a test you take alone. In practice it is a contest: one rival holds the slot on specific questions. No checklist tells you which item matters for which question. The answer that recommends the rival does.
So the work has four steps:
- Find the buyer questions where a named rival gets the recommendation and you are absent or only mentioned.
- Read why: which condition of the question the rival meets, which sentence gives the reason, which source stands behind it.
- Close the one gap with the most engines lost and the most value to your buyers.
- Check by asking the same questions again, and count the switch only when it is larger than normal variation.
Is it risky to ignore how competitors show up in AI answers?
Yes, when the pattern is real. If most engines recommend the same rival on a question your buyers ask right before they choose, that rival’s name is the answer those buyers get. Nothing on your own website tells you this is happening. You only see it by reading the answers.
The opposite risk is just as real: a screenshot of one answer is a snapshot. Answers vary from run to run, and a rival recommended once today may be gone tomorrow. Rewriting pages because of one answer is work on noise. Read the risk from a pattern across questions, engines and weeks, and act on that.
No tool or tactic can guarantee that an engine switches to you. What you control is whether the pages it reads give it a reason to.
Find the questions you lose to one named rival
Your first job is a short list: the buyer questions where one named rival gets the recommendation. Two kinds of loss belong on it, and they call for different work.
- Absent: the rival is named and you never are. The pages the engine reads do not connect you with the question at all.
- Named, not picked: you are in the answer and the recommendation goes to the rival. The answer has a reason to prefer them, and that reason is usually in the text.
Ask the questions without your brand name, the way buyers ask them, and ask the same ones in every engine. Ask each one at least three times per engine, on different days, and put a rival on the list only when it is the pick in most of those runs. Ask logged out, or in a new chat with memory turned off, the way a new buyer would. A question with your name already contains you and tells you nothing about the open comparison. How to build a question set that mirrors real buyers is in our guide to prompt tracking. For each answer, note three things: who is named, who gets the recommendation, and which page the answer cites for it.
Check your list of rivals too. The rival taking your recommendation is often not the one your sales team names. FameScan lists brands that appear in at least two stored answers and that you do not track yet, under Competitors, Brand suggestions. One click on “Track as competitor” adds them.
You do not have to keep the list of lost questions by hand. The Action Center shows “Competitors own 3 of your tracked questions” when, across at least two answers to a question, rivals are named and you never are; each evidence line names the question and the rival it goes to. When you are named in at least three answers but actively recommended in under 40 percent of them, it shows “Named often, recommended rarely”, with evidence lines that name the rival the answer picked instead, where there was one. Every line opens the stored answers behind it.
Mentioned, cited, recommended: which win did the rival get?
A mention is not a citation, and a citation is not a recommendation. They are three different wins, and mixing them hides the real problem. So ask first which of the three your rival won.
| Win | In the answer | FameScan metric |
|---|---|---|
| Mention | The brand name is in the text | Visibility |
| Citation | A page of the brand is linked as a source | Citation rate |
| Recommendation | The answer tells the reader to choose that brand | Recommendation rate |
The work follows from the win. If the rival is cited and you are not, the job is sources; which source to work on first is in our guide to AI citation tracking. If you are both named and only the rival is recommended, the job is the reason: the fact, the comparison, the review that makes the engine choose. This guide is about that third win.
Visibility and citation rate count the same answers to non-branded questions: those that name you, and those that cite your domain. Recommendation rate in FameScan counts only answers to non-branded questions that name you and that FameScan could judge: in what share of them does the engine actively make you its pick, rather than just list you? Like visibility and citation rate, it carries a 95% interval and a verdict. FameScan shows your share of voice against your rivals next to it, as a descriptive number; how it is calculated is in our guide to AI share of voice.
Why does ChatGPT recommend your competitor instead of you?
Usually because the pages ChatGPT read for that question answer it with your rival’s name and a reason, and do not do that for you. The same holds for Perplexity, Claude and Google AI Overviews, which also answer from pages they find. If the answer shows no sources, there is no cited page to work on; the pick shows which brands the engine already links with the question. The stored answer shows which case you have.
Read one answer in three passes:
- The condition: which part of the question does the pick serve? A use case, a team size, a price limit, a location, an integration.
- The reason: in what words does the answer justify the pick? That reason is the fact you are missing, or never state plainly.
- The source: which cited page carries that reason? The rival’s own page, a comparison or best-of list, a review profile, a forum thread.
Where each engine reads before it recommends differs. The table sums up what the vendors themselves say, as of November 14, 2026.
| Engine | What it reads |
|---|---|
| ChatGPT | Searches the web for its search answers. According to OpenAI, sites that block OAI-SearchBot are not shown in ChatGPT search answers, apart from plain navigational links. |
| Perplexity | Searches the internet; each answer has numbered citations that link to the original sources. |
| Claude | Uses a search tool on the live web when it helps; its search answers include citations. |
| Google AI Overviews | Shows pages that are indexed and eligible to appear with a snippet, with no additional requirements; may run several related searches for one answer. |
In FameScan, Gemini answers from what it already knows, without a live search. So read its pick next to the others: it shows which brands Gemini already links with the question, while the source work happens through the four engines that search. Before you change content there, check that crawlers may read your page at all; how that works for ChatGPT is in our guide to ranking in ChatGPT.
Which competitor gap do you close first?
Close first the gap where one rival gets the recommendation on the most engines, on a question close to a purchase, with a reason you can truthfully meet. Four checks sort your list:
- Engines: on how many of the five is the rival the pick? Three of five beats one of five.
- Closeness to a purchase: “best software for …” or “X or Y for …” sits nearer the decision than “what is …”. That is your call, best made with your sales team.
- A reason you have: if the answer justifies the pick with a fact that is also true of you, you can close it. If the rival really has something you lack, that is a product question, not a content task; take the next question.
- Sources in the answer: a pick backed by cited pages can move when those pages change. A pick in an answer without sources, as with Gemini here, rests on what the engine already knows; take the question with sources first.
On a tie, take the question where you are already named. The engine already has you on its list there; what is missing is the reason, not awareness. The table maps the most common reasons to their work.
| What the answer shows | Your fix |
|---|---|
| It repeats the rival’s fact and cites the rival’s page | State the fact plainly on your page for that use case Product marketing |
| A third-party best-of list is among the sources | Send the editor the facts the list compares on PR or content |
| Capterra, G2 or a Google profile is cited | Complete your listing, ask customers for reviews Customer success |
| You are absent in the engines that search; your domain is never cited | Check crawler access and text without JavaScript Web team |
| The answer gives an old price or feature of yours | Get the cited source corrected Source owner |
Write the fact the way the answer’s reason sentence puts it, in one plain sentence near the top of your page for that use case. Add the detail the question asks for: which insurers, which integration, up to which price.
For the second and third rows, the gap analysis in our guide to AI citation tracking helps; for the fourth, the AI readiness audit checks robots.txt for each AI crawler and whether your key pages show their text without JavaScript. Which on-site steps come first in general is in our guide to generative engine optimization.
A worked example: one clinic question, five engines
The Action Center shows “Named often, recommended rarely”. One evidence line names the question “best scheduling software for physical therapy clinics” and says the AI picked Movara instead. Slotfin opens the stored answers for that question:
- ChatGPT recommends Movara “because it handles insurance billing”, names Slotfin further down, and cites Movara’s feature page and a Capterra category page.
- Perplexity recommends Movara for the same reason and cites a best-of list on a clinic trade site. Slotfin is absent.
- Google AI Overviews recommends Movara first, for insurance billing, and links the same best-of list.
- Claude names Slotfin and Movara and recommends Slotfin for online booking.
- Gemini names Movara and Pracsis, without sources.
The pattern is clear: Movara is the pick on three of five engines, and all three give the same reason. Slotfin handles insurance billing too. But that is only in a help article, not on the page for physical therapy clinics and not in its Capterra listing. Slotfin meets the condition; no page the engines read says so.
Slotfin picks this question over its two other lost ones: three engines, close to a purchase, a reason Slotfin truthfully meets. Product marketing writes one plain paragraph on insurance billing for the clinic page, customer success updates the Capterra listing, and the best-of list’s editor gets the facts. In the Action Center, Slotfin clicks “Start working” on the item.
Two weeks later the item’s “Impact since start” line shows the recommendation rate in the 14 days before and after. It went up, and the verdict reads “within noise”. Slotfin reports it as “change live, no proven switch yet” and reads the same questions again next period.
How to see the switch in FameScan
In FameScan, every step of this work has a fixed place, and every number traces back to the stored answers behind it:
- Competitors: Share of voice, Category comparison and “Who wins on which model” show against whom you lose, and where. Clicking a rival’s “Mentioned in” count shows the prompts that mention it. More on the competitor comparison page.
- Prompts: the Recommended column shows per question how often you are the pick. Expand a lost question and click “Ask the Copilot”: it analyzes who the engines recommend instead and why, which sources they cite, and drafts a short plan.
- Copilot: ask “Compare our recommendation rate with Movara’s over the last 30 days, with intervals.” It answers from the same data, with charts. If you work in Claude, the MCP server returns the same competitor comparison, and its ready-made “Win a prompt” prompt explains one lost question.
- Overview: the Recommendation rate tile shows the value, its 95% interval and the verdict for the selected period against the one of equal length before it. The Real changes list also shows statistically real moves of your tracked rivals.
- Alerts: an alert tells you when AI answers recommend your brand more often, once your recommendation rate really rises.
- Impact: when you set “Named often, recommended rarely” to started or completed, FameScan compares your recommendation rate in the 14 days before and after. A verdict appears after at least 7 days and 20 answers that name you; until then it says it is too early to judge.
Your next 10 minutes: one gap card
Take one buyer question close to a purchase where you suspect a rival wins, ask it without your brand name, logged out or with memory off, three times in each of two engines, and fill in one card for that question alone. It fits in a message to your team:
One card, one question, one rival. Two weeks later, ask the same question the same number of times: if you are now the pick in most runs where the rival was, for the same reason, you have a first sign. The switch counts when the verdict on your recommendation rate reads Real change.
Sent as it is, the card is also your status report for a boss or client: the question, the rival, the stored answers that recommend it in their own words, the change and its owner, and the recommendation rate with its interval and verdict. Every line traces back to an answer.
Frequently asked
How do I get ChatGPT to recommend my business?
Start with the buyer questions where ChatGPT already recommends someone else. Read those answers: which condition of the question does the rival meet, and which page backs it? Then state that fact plainly on your own page and make sure the cited sources carry it. Ask the same questions again later. One answer can change by chance, so judge the switch on a set of answers, not a single screenshot.
Why does AI recommend my competitor and not me?
Usually because the pages the engine reads for that question answer it with your rival’s name and a reason, and not with yours. Their page states the condition the question asks about, a comparison or review page lists them, or your page leaves the fact out. The stored answer and its sources show which.
Can I pay to be recommended by AI?
Paid ads are a separate channel from the organic answer: report them separately and never count an ad as a recommendation. The organic pick comes from the pages an engine reads and from what it already knows, so the work is getting those pages to name you for the question and give a reason.
Is being mentioned the same as being recommended?
No. A mention names you, a citation uses a page as a source, a recommendation makes you the pick. You can be mentioned in every answer and still lose the recommendation to a rival. Track the three separately, on the same questions, or you will not see which work is missing.
How long until AI recommends us instead of a competitor?
No one can promise a date or a result. Engines re-read sources at their own pace, and answers vary between runs. Ask the same questions in the same engines after the change. Count the switch only when the change in your recommendation rate is statistically real against the period before, not when one answer flips.
Which competitors should we compare against?
The ones the engines actually name for your buyer questions, not only the rivals your sales team sells against. FameScan suggests brands that appear in at least two stored answers and that you do not track yet. Add the ones that keep showing up on the questions you lose.
Sources & further reading
Sources checked on 14 November 2026.




