How to get ChatGPT to recommend your business

We ran 60 scans of our own, across 41 sites in 44 categories. They produced 45,309 citations from 7,260 AI answers, and one pattern stood out: listicles and roundups supplied 67.5% of citations, while the scanned product's own site supplied 1.5%. The useful question is therefore less about persuading ChatGPT and more about becoming a clear, credible source in the places it already uses.

There is no form, and roundups dominate

Those answers cited 7,624 distinct domains and 16,506 distinct pages. A business may be named because a page an engine reads includes it, because its own site answers a buyer's question, or because several sources describe the same thing in compatible ways. There is no general form where you enter your company name and wait for a recommendation to appear.

OpenAI says it uses web crawlers and user agents for its products, either automatically or when a user requests it, in its documentation about bots and crawlers. It also says that OAI SearchBot is used to surface websites in ChatGPT search features. That makes crawl access a condition of entry, not a recommendation in itself, and the wider citation data shows why access alone is not enough.

Across the full scan, which includes Perplexity, Gemini, the ChatGPT family and Claude, listicles and roundups accounted for 67.5% of citations, or 30,585 citations. Competitor product pages added 24.4%, or 11,052 citations. Together, those two page types accounted for 91.9% of citations, leaving every other type of page to share the remainder.

We cannot treat that page-type mix as a ChatGPT-only breakdown. ChatGPT-family answers in our sample cited fewer sources, 3.4 per answer on average, and named the scanned product in 4.4% of answers. Those figures suggest a similar dependence on a small set of available sources, but they do not prove that every ChatGPT recommendation follows the same route.

Your own product page still matters. It gives other writers something accurate to check, and it gives a crawler a clear description of what you sell. But the data says that relying on your homepage alone is a poor plan. In the 120 sites we checked, only 5 passed the check for their own pages getting cited.

The figure shows the gap plainly. Roundups are where most of the recorded citations sit, while a product's own site is a small part of the sample. That does not mean every business needs a large public relations campaign. It means you should look for the specific pages that answer the same buyer questions as your product.

1. Build a list of pages worth earning

Start with buyer language, not your brand name. Gather the questions people ask before they know which company they want, then search for pages that already answer those questions. Look for comparisons, alternatives, category roundups and practical guides where your business would be a sensible inclusion.

Read each page as an editor would. Does it list products by use case, price, company size or location? Does it explain why a business belongs on the list? Does it link to the businesses it names? Those details tell you whether a page is a realistic target, rather than a famous URL that has no room for another entry.

Then check the pages that appear in AI answers for your category. Our weekly research on who owns AI answers shows the kind of source pattern we measure, while the free AI visibility checker lets you test questions for a specific business. The aim is not to collect mentions for their own sake. It is to find the pages that sit between a buyer's question and a recommendation.

Contact the editor only when you have a useful reason. Send a correction, a missing price, a new comparison point or evidence that your business serves the stated audience. A generic pitch gives an editor no clear reason to act on it.

2. Make your site easy to quote

A writer or crawler should be able to answer basic questions from the page itself. What is the product? Who is it for? What does it cost? What does it replace or beat, and under what conditions? Put those answers in ordinary text near the relevant product or service, rather than making the reader reconstruct them from slogans, screenshots and a pricing calculator.

Our readiness checks found this was the weakest area. A plain statement covering what the business is, who it serves, its price and its alternative passed on 20 of 120 sites, failed on 43 and warned on 55. By contrast, a homepage answered a plain request on 118 of 120 sites. Being reachable is common. Being specific is harder.

Use one clear H1, descriptive page titles and a page for each important comparison or use case. Answer the questions a buyer would ask before contacting sales. If your price varies, explain what causes the variation and give the reader a useful starting point. This creates text that can be understood, checked and quoted without a private product demonstration.

Structured data helps machines identify parts of a page, but it cannot rescue vague copy. Organization schema passed on 72 of 120 sites, while Product or SoftwareApplication schema passed on 60. Add the markup that accurately describes your business, then make sure the visible page says the same thing.

3. Remove the technical reasons to disappear

Check robots.txt first. The readiness scan found that AI crawlers were allowed on 115 of 120 sites, so this is not the only problem, but the 5 failures are still avoidable. OpenAI says that sites opted out of OAI SearchBot will not be shown in ChatGPT search answers, although they can still appear as navigational links. It recommends allowing OAI SearchBot and requests from its published IP ranges if you want your site to appear in search results. Treat that as a crawl requirement, not a promise of a recommendation.

Keep the important copy in the page output without requiring JavaScript to reveal it. The content check passed on 91 of 120 sites, failed on 18 and warned on 9. Publish a sitemap, use canonical and Open Graph tags, and link the useful pages from the site so a crawler does not have to guess which URLs matter. The same OpenAI page says that OAI SearchBot and GPTBot settings are independent, and that a robots.txt update can take about 24 hours to adjust for search results.

The free AI readiness check covers these foundations, including crawler access, rendering, page structure and structured data. It will not earn a roundup placement for you. It can show where your site is making the job unnecessarily difficult.

4. Publish the comparisons buyers need

Roundups are not the only useful pages to create. Comparison and alternatives pages give your site a chance to explain where your product fits, especially when a buyer already knows a competing category or tool. In the readiness sample, comparison or alternatives pages passed on 54 of 120 sites, failed on 20 and warned on 46.

Write these pages for a real decision. State who should choose each option, where the option falls short, what it costs when that information is available, and what kind of business it suits. Include competitors fairly. A page that calls every alternative inferior gives other sources little reason to trust it, and gives readers little reason to stay.

Questions answered on the page were another weak area. That check passed on 56 of 120 sites and warned on 62. Use the questions your sales team hears, support tickets contain or buyers type into search, then answer them in language that can stand alone when quoted by an AI system.

The comparison check passed on 54 of 120 sites, so this is a practical gap rather than an abstract content exercise. A useful page gives a writer or an AI system a clear account of fit, price and tradeoffs, which makes your business easier to verify against alternatives.

A worked example with the numbers in view

Suppose you sell a project management product and want ChatGPT to recommend it for small teams. Start by mapping the buyer's question to the pages and checks that can support an answer, rather than assuming the product page will do all the work.

First, find relevant small team software roundups and comparison pages. Read their inclusion rules, note which claims they need to verify, and prepare accurate information about team size, pricing, setup and use cases.

Second, create an alternatives page and a comparison page on your own site. In the readiness sample, comparison or alternatives pages passed on 54 of 120 sites, failed on 20 and warned on 46. That gives you a concrete gap to inspect, not a guaranteed route into an answer.

Third, check the technical basics. In the 120 site readiness sample, 115 allowed AI crawlers, 100 had one clear H1, 90 made a price visible or linked to a pricing page, and 91 had content present without JavaScript. These are reference points from our sample, not a target score handed down by ChatGPT.

Finally, test the buyer question itself. ChatGPT-family answers in our sample cited 3.4 sources per answer on average, and named the scanned product in 4.4% of answers. Your goal is not to force a mention in every answer. It is to become one of the credible, easy to verify sources available for the questions your product genuinely answers.

What this sample leaves unanswered

We cannot tell you that a particular page will make ChatGPT recommend your business. Our data measures samples: 45,309 citations across 7,260 AI answers, from 60 scans of 41 sites in 44 categories. It does not measure every ChatGPT answer, every search result or every business, and it does not prove that one change caused one recommendation.

We also cannot tell you how long a new mention will take to appear, whether a specific editor will add your company, or which page ChatGPT will select for a particular user. The source mix varies by question and category. We are showing you the edges of what 60 scans across 44 categories can tell you, nothing beyond that.

What we can say is narrower and more useful. There is no submission form to rely on. Most of the citations we recorded came from listicles and roundups, with much of the rest coming from competitor product pages. Make your own site clear, allow the relevant crawler, and work to become a factual inclusion on the pages that already shape the category.

We opened with 45,309 citations and a missing form. The form is still missing. What has changed is the shape of the work: you now have a source pattern to investigate, a site to make legible and a set of pages where a recommendation can actually begin.