How do you get ChatGPT to recommend your product?
ChatGPT recommends the products that the pages it reads recommend. When a buyer asks ChatGPT or Perplexity for the best tool in your category, the engine searches the web, reads a handful of pages and writes its shortlist from them. In our citation data, 89% of those pages were not the vendors' own sites. So the work is: find the buyer questions that matter, see which pages the answers cite, get onto those pages with an accurate description, make your own site state the same facts plainly, let the AI crawlers in, and retest. The steps below show how.
768 citations from 64 AI answers about B2B software.
We asked an OpenAI model with web search and Perplexity Sonar the same buying questions about project management and email marketing software, twice each, and recorded every cited URL. Full method and limits are in the citation study.
11%
of cited URLs were the vendors' own websites
Only 81 of 768 cited URLs pointed to the sites of the vendors being compared. The other 89% were comparison articles, "best tools" round-ups, reviews and other companies' blogs, such as custify.com, ventureharbour.com, project-management.com, sequenzy.com, encharge.io.
2 of 64
answers cited G2, Capterra or another review marketplace
Review marketplaces were almost absent as direct sources, and both citations were to G2's editorial blog rather than a product review page. Reviews may still matter indirectly when round-up articles quote them.
28 of 32
Perplexity Sonar answers cited Reddit
The OpenAI model cited Reddit in 0 of 32 answers to the same questions. Where you need to be present depends on which engine your buyers use.
55%
average source overlap when the OpenAI model answered the same question twice
Asked twice within minutes, it cited an identical set of domains in only 2 of 16 question pairs. One screenshot is not a baseline; ask every question more than once.
Limits: a pilot sample of two categories, collected through the engines' APIs with the same retrieval layer, not the consumer ChatGPT app. Treat the numbers as a direction to test in your own category, not a rule.
Seven steps to get your SaaS into ChatGPT recommendations.
- 01
Write down the buyer questions that decide your category.
List 10–30 questions a buyer would type before a shortlist exists: “best [category] for [segment]”, “[competitor] alternatives”, “[category] that integrates with [tool]”, “[category] pricing for a 50-person team”. Unbranded questions matter most; nobody who already knows your name needs a recommendation.
- 02
Ask them where your buyers ask, and ask twice.
Run each question in ChatGPT with search on, Perplexity, Gemini and Google AI Mode. Record which products are recommended, in what order, and every URL shown as a source. Repeat the run: if the recommendation changes between runs, you have learned how unstable the answer is before spending anything.
- 03
List the third-party pages the answers cite.
In our data most cited pages were not vendor sites. For each cited round-up, comparison or review article, check three things: are you on it, is your description correct, and is your category label the one buyers use. These pages are your real distribution list.
- 04
Get onto those pages accurately.
Contact the editors of the cited articles with a short, factual pitch: what you do, for whom, price, one proof point, and a screenshot or trial access. Correct wrong descriptions. Earn inclusion rather than buying it; paid placements that read as ads are rarely the pages answer engines cite.
- 05
Make your own site answer the same questions plainly.
Engines can only repeat facts that exist somewhere. Publish a page per buyer question you want to win: who the product is for, a public pricing page, an integrations page, honest comparison and alternatives pages, and use-case pages. Use the category words buyers use, not internal positioning language.
- 06
Let the answer engines read you.
Check that robots.txt does not block the search crawlers answer engines use, such as OAI-SearchBot and PerplexityBot. OpenAI says sites that opt out of OAI-SearchBot are not shown in ChatGPT search answers. Make sure key facts are in the HTML, not only inside images or client-side scripts.
- 07
Retest the same questions after the change.
Re-run the frozen question set four to eight weeks after the fixes go live, under the same conditions and with the same number of repeats. Compare recommendations and cited pages question by question rather than watching a single blended score.
Step six takes a minute with the free AI crawler checker. For steps one to three on five questions, run the free AI visibility audit; to run the whole sequence yourself, follow the AEO audit checklist.
Do review sites and Reddit get you recommended?
Review marketplaces were cited in 2 of 64 answers in our sample, so a G2 profile on its own is unlikely to be the page an answer engine quotes. Keep it accurate anyway: the round-up articles that are cited often repeat review scores and positioning. Reddit is engine-specific. Perplexity Sonar cited it in 28 of 32 answers and the OpenAI model in 0. If your buyers use Perplexity, genuine, disclosed participation in the threads where your category is discussed is worth the time. Fake posts are not.
What does not get you recommended
Only adding schema or an llms.txt file
Both can help machines parse a site. Neither makes an engine choose you when the pages it reads recommend someone else.
Rewriting every page “for AI”
If third-party pages decide most answers, polishing owned copy alone changes little. Fix owned facts, then work on the sources.
Judging progress from one screenshot
Answers vary between runs and engines. Decide from repeated runs of a fixed question set.
Buying reviews or fake community posts
It breaks platform rules, and fabricated praise is the kind of signal that gets removed or ignored.
See which questions you lose, to whom, and which pages decide them.
The free self-check gives you five buyer questions to ask and tallies which competitors are recommended instead of you. The paid AI visibility audit covers up to 30 questions across three engines, compares you with three competitors and ends in a prioritized 90-day plan. Neither guarantees a recommendation; both tell you where to spend effort first.
Questions SaaS teams ask
How do I get ChatGPT to recommend my company?
Find the buyer questions where you should be recommended, see which pages ChatGPT cites when it answers them, get accurately included on those pages, publish clear owned pages for the same questions, make sure OAI-SearchBot can crawl your site, and retest. There is no form to submit and no way to buy an organic recommendation.
Do G2 and Capterra reviews help you get recommended by ChatGPT?
In our sample of 64 AI answers about B2B software, review marketplaces were cited in 2. That does not make reviews useless: comparison articles that engines do cite often summarize review scores. Keep your profiles accurate, but do not expect a review page alone to change answers.
Does ranking on Google mean ChatGPT will recommend us?
Not necessarily. A SaaS can rank for its own name and still be missing from unbranded category answers, because those answers lean on round-up and comparison pages rather than your homepage.
How long does it take to show up in ChatGPT answers?
It depends on when the cited pages change and when the engine retrieves them again. Plan to retest four to eight weeks after a change and treat earlier movement as noise unless it repeats.