How to Show Up in ChatGPT Recommendations

How does ChatGPT decide what to recommend?

When a user asks ChatGPT for a recommendation, the model often runs a live web search, retrieves pages, and assembles an answer from the passages it can read cleanly. It is not reading your whole site. It pulls discrete chunks of content, and Semrush's analysis of AI search behavior found that models prioritize relevance and information quality in those chunks over page experience metrics like load speed.

This retrieval step is why AI visibility differs from classic rankings. In the same Semrush study, ChatGPT cited pages ranking at position 21 or lower in traditional search almost 90 percent of the time. A practice that never cracked page one for a competitive term can still be the source ChatGPT quotes, if the page answers the question directly.

The audience is worth the effort. Semrush found the average AI search visitor is 4.4 times as valuable as the average traditional organic search visit, measured by conversion rate, because these visitors arrive having already compared options.

Let OpenAI's crawlers in first

None of the content work matters if the crawler is blocked. OpenAI documents four crawlers, and the one that governs search visibility is OAI-SearchBot. Per OpenAI's own documentation, sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers, though they can still appear as navigational links.

The crawlers are controlled independently in robots.txt, so you can allow OAI-SearchBot for visibility while making a separate decision about GPTBot, which collects content for model training. Changes take roughly 24 hours to register on OpenAI's side.

Check three things this week: your robots.txt does not disallow OAI-SearchBot, your firewall or CDN is not returning 403 errors to it, and your sitemap is current.

• OAI-SearchBot: powers ChatGPT search results, required for recommendation visibility
• GPTBot: collects content for training, a separate opt-in decision
• ChatGPT-User: fetches pages when a user asks about your site directly

Structure pages so answers are extractable

Because the model works in passages, each service page should contain at least one self-contained answer: a plain question as a heading, followed by a direct 40 to 60 word response that names your practice, the service, and the location. Bury the answer in a paragraph about your philosophy of care and the model has nothing clean to lift.

Extractability also means the answer exists in the initial HTML. Content hidden behind accordions that only render on click, or injected by scripts after load, is unreliable for retrieval. State prices, conditions treated, appointment logistics, and credentials as concrete facts, not adjectives.

Keep in mind that people phrase AI queries conversationally. Semrush's clickstream analysis found 65 to 85 percent of ChatGPT prompts do not match traditional search keywords. Write headings the way a patient would actually ask the question.

Keep your entity information consistent everywhere

Language models corroborate. Before recommending a clinic, the system cross-references your name, address, phone, services, and credentials across your site, Google Business Profile, directories, and review platforms. Conflicting details lower confidence, and a low-confidence entity gets omitted rather than recommended.

Treat your practice as one entity with one canonical record. The exact same business name, the same address format, the same service list, and current hours, replicated everywhere you appear. Structured data helps here: LocalBusiness or MedicalBusiness schema on your site gives engines a machine-readable version of those facts to match against.

Audit quarterly. A moved office or a renamed service line that only gets updated on your website creates precisely the contradiction that suppresses recommendations.

Earn presence on the sources engines already trust

Citations concentrate. Semrush's traffic analysis found over 30 percent of all ChatGPT referral traffic goes to just 10 domains, and its citation research shows roughly half of ChatGPT's links point to business and service sites, with the rest spread across news, forums, directories, and community platforms like Reddit and Quora.

For a practice, this means your own site is necessary but not sufficient. Reviews on major platforms, listings in health directories, local press mentions, and genuine answers in community threads all create third-party corroboration the model can retrieve.

Traditional SEO still feeds the machine. Semrush notes that the brands winning AI visibility are the same ones seeing strong SEO results, and Google itself receives over 21 percent of ChatGPT's referral traffic. The two channels reinforce each other rather than compete.

Measure share of voice, not just rankings

AI visibility needs its own scoreboard. The practical unit is share of voice: for a defined set of prompts a prospective patient would ask, what percentage of responses mention or cite you versus competitors. Run the same 15 to 25 prompts monthly, in fresh sessions, and log who appears.

Track three numbers over time: mention rate (how often you appear at all), citation rate (how often your site is the linked source), and referral traffic from AI platforms in your analytics. The opportunity is growing; Semrush measured a 206 percent increase in ChatGPT referral traffic between January 2025 and January 2026, with traffic spread across roughly 170,000 domains by February 2026, up from about 71,000 in late 2024.

This is the discipline Rank and Rejuvenate applies for wellness practices: crawler access verified, answers made extractable, entity records reconciled, and share of voice tracked monthly, so AI visibility shows up as booked appointments rather than a line in a marketing report.