How to get ChatGPT to recommend your product

A founder is between meetings, checking whether a buyer will find the product before the next call. They type, “Which tools should I consider?” and ChatGPT returns a short list.

The comparison target

Assume, as this piece does, that a business can be recommended on reputation alone. A product usually needs a clearer case: what it does, who it suits, what it costs and how it differs from the alternatives.

You are not persuading ChatGPT with a slogan. You are making the pages it reads specific enough to support a buying decision. That means clear product facts, useful comparisons and evidence outside your own site.

Start with the pages that already influence answers.

Citation paths

The weekly research on who owns AI answers records 45,309 citations across 7,260 AI answers. Those answers came from 60 scans of 41 sites in 44 categories. Listicles and roundups supplied 67.5% of citations, or 30,585 citations. Competitor product pages supplied 24.4%, or 11,052 citations. Together, those page types supplied 91.9% of recorded citations.

Find the category roundups, alternatives pages and competitor comparisons that buyers already use. Then help those pages describe your product accurately. An independent comparison can give ChatGPT more useful evidence than another page that describes your own features in isolation.

Your homepage still matters. It is rarely the whole recommendation path.

Product facts

A comparison needs stable facts. State the category, intended customer, main use case, price, integrations, setup requirements, limits and meaningful alternatives in language a reader can reuse. “The leading platform for modern teams” gives little to compare. “A reporting tool for small agencies, priced from the published rate, for teams that need this workflow rather than that one” gives a source something more precise.

The free AI readiness check audited 120 sites. 20 passed the check for a plain statement of what the product is, for whom, at what price and against what. 43 failed and 55 warned. By comparison, 92 passed the check that their title and description said what the product is.

Here is a site-shaped example using those results. Imagine a site with no plain product statement, no comparison page and full access for AI crawlers. The site may be findable, but a listicle writer still has to work out the category, buyer, price and alternative before placing it in a recommendation.

The arithmetic shows the gap. 20 of 120 sites passed the plain statement check, which is 1 in 6. So roughly 5 out of 6 sites in this audit did not give a comparison writer a clean product-and-position sentence, either because they failed or because the audit warned about the information.

Write that sentence once, put it on the homepage and the comparison page, and stop there.

Alternative pages

Buyers arrive with a job, a budget, an incumbent tool or a shortlist. Build pages around those decisions. A page about your alternatives can help when it explains fit. So can a page comparing your product with a named competitor, a category roundup or a guide for a buyer with a clear constraint.

The page needs more than a name swap. Explain who each option suits, what the options share, where they differ, what each costs when the facts support a comparison, and what a buyer gives up by choosing one. Put the direct answer near the top, then show the reasoning.

The readiness audit found that 54 of 120 sites passed the comparison or alternatives page check, while 20 failed and 46 warned. Name the two kinds of buyer who should pick you, and the one who should not.

CiteGraph’s methodology page reports that, of the 40 domains cited three or more times in its first scan, 34 were still cited 17 days later. Of 179 exact pages cited in that scan, 41 were cited again. Those figures describe repeated citations in one project, not a permanent ranking. The methodology page explains the measurement.

The source ledger contains no verified pull quote from a vendor’s own page, so there is no vendor quote here. Better an empty quote slot than invented authority.

Technical access

Crawlability will not make a weak comparison strong, but blocked or unclear pages can stop useful evidence being found. In the audit of 120 sites, 115 passed the check for AI crawlers allowed in robots.txt. 99 had a published sitemap. 91 had content present without JavaScript. 91 passed canonical and Open Graph checks, while 27 warned.

Keep important product facts in the rendered page. Link the pricing page. Give each page one clear job. Keep canonical signals consistent. Use structured data when it matches the visible content.

Organization schema passed on 72 of 120 sites. Product or SoftwareApplication schema passed on 60, while 58 failed. Markup cannot supply facts the page does not state.

Questions are another practical layer. 56 of 120 sites passed the check for questions answered on the page, while 62 warned. Answer who the product is for, what it replaces, how it differs, what it costs and when it is a poor fit.

Use the free AI visibility checker to turn these issues into a work queue, not to predict a particular ChatGPT answer.

The scan recorded 7,624 distinct domains and 16,506 distinct pages. Community threads supplied 3.4% of citations, or 1,534 citations. Review platforms supplied 0.8%, or 348 citations. These shares are smaller than those of roundups and competitor pages, but they can add context to a recommendation.

Put the price and the comparison table in one page you update, not three.

Measurement and upkeep

Our numbers describe 442 buyer questions across 44 categories, using 60 scans of 41 sites. The sample measures selected questions and sources. We print those limits; we do not know what we have not measured.

Answers also move. In the CiteGraph project using scans 17 days apart, 34 of 40 domains cited at least three times in the first scan were cited later. Across every cited domain, 69 of 102 came back. Exact pages were less stable, with 41 of 179 cited again. The methodology page sets out how those measurements were made.

Track the buyer questions that matter, the products named beside yours, the pages cited and the facts that appear in the answer. Check again after a meaningful product, pricing or competitor change. Watch whether your product enters the comparison set, rather than whether one prompt produces a flattering sentence.

The founder between meetings can ask a better question now. Not how to make ChatGPT name the product, but which comparison pages can place it accurately, and whether those pages have enough evidence to do the job.