How to get your brand recommended by ChatGPT, Perplexity and Claude
ChatGPT SEO explained: how assistants choose which brands to recommend, how to track your visibility, and eight steps to earn a place on the shortlist.
Key takeaways
- Assistants recommend brands that are consistently described as credible options across many independent sources.
- Live retrieval means recent, well-ranked comparison and review content directly affects answers.
- You can measure this: run a fixed prompt set monthly and record who is named.
- Accuracy matters as much as presence; fix wrong descriptions of your product.
To be recommended by ChatGPT, Perplexity or Claude, your brand needs to be described as a credible option in the sources those assistants learn from and retrieve. That means well-ranked comparison and review content, consistent entity information, coverage in trusted publications and a site the crawlers can read. This is sometimes called ChatGPT SEO or LLM SEO; we call it answer engine optimization.
Here is how assistants build a shortlist, how to measure whether you are on it, and eight steps to get there.
How assistants decide who to recommend
When someone asks "what is the best project management tool for agencies", the assistant does two things. It draws on what it learned during training about which products exist and how they are described, and, in most modern assistants, it runs a live search and reads several pages before answering. The shortlist reflects both.
- Training data favours brands with broad, consistent coverage over years: reviews, articles, forum discussions, documentation.
- Live retrieval favours pages that rank for the query right now, especially comparison and "best of" content.
- Description accuracy depends on how clearly your own site and third parties state what you do, for whom and at what price.
The practical consequence is that you cannot optimise a single page and expect to appear. You influence a body of evidence.
Measure it before you change anything
Build a prompt set of twenty to fifty questions your buyers ask, drawn from sales calls and search data. Run each prompt monthly in ChatGPT, Perplexity, Claude, Gemini and Copilot. Record which brands are named, which sources are cited, and whether the description of your product is accurate. Your share of model is the percentage of answers that mention you. Our free AI Visibility Checker runs this for five prompts to get you started.
Eight steps to earn the recommendation
1. Make your site readable to AI crawlers
Allow GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot and Google-Extended in robots.txt unless you have a specific reason not to. Publish an llms.txt file that summarises your key pages in plain language. Keep product information in HTML rather than images or scripts.
2. State plainly what you are
Your homepage and about page should answer, in one paragraph, what you do, for whom, where and how you are different. Repeat that description consistently on social profiles, directories and press boilerplates. Inconsistent descriptions produce inconsistent answers.
3. Publish comparison and alternatives content
"X versus Y" and "best X for Y" pages are among the most retrieved for recommendation prompts. Write them honestly, with real criteria and tables, including competitors. Honest comparison content is cited; one-sided content is not.
4. Earn reviews on the platforms models trust
G2, Capterra, Trustpilot, Google reviews and category-specific review sites are heavily represented in training data and retrieval. A steady flow of detailed, authentic reviews shapes how assistants describe you.
5. Get covered where your category is discussed
Industry publications, newsletters, podcasts and community threads all feed the evidence base. A digital PR programme aimed at these outlets improves both rankings and recommendations.
6. Add structured data and author identity
Organization schema with sameAs links, Product or Service schema, and author pages with Person schema help systems connect your content to a real entity.
7. Rank for the underlying keywords
Live retrieval pulls from search results. If you do not rank for "best X for Y", you will rarely be read when that prompt is asked. Classic SEO remains the entry ticket.
8. Fix inaccuracies
When an assistant describes you wrongly, trace the claim to its source, correct the source where you can, and publish clear, current information on your own site. Accuracy problems usually come from outdated third-party pages.
Which sources each assistant relies on
The engines are not identical, and the differences shape where to focus.
| Assistant | Leans on | Practical implication |
|---|---|---|
| ChatGPT | Training knowledge plus web search for current questions | Broad, consistent coverage over time; comparison pages for live lookups |
| Perplexity | Live retrieval with visible citations | Rank for the query; be on the pages it cites, often listicles and reviews |
| Claude | Training knowledge plus optional web search | Entity clarity and independent coverage |
| Gemini | Google's index and knowledge graph | Classic SEO strength, structured data, Google Business Profile |
| Copilot | Bing's index | Do not neglect Bing Webmaster Tools and indexing |
What not to do
- Do not stuff pages with "best" claims about yourself. Assistants weight independent sources more than self-description.
- Do not fabricate reviews or pay for placements in listicles that disclose nothing. Both are detectable and damage trust.
- Do not block crawlers and expect better control. You get less accuracy, not less mention.
How long it takes
Retrieval-driven improvements, such as ranking a strong comparison page, can change answers within weeks. Training-data-driven improvements compound over months as new model versions incorporate the evidence you have built. Plan for a six to twelve month programme and measure monthly.
If you would like a benchmark of where you stand across every major assistant, start with a free AI SEO audit.
Frequently asked questions
Can you pay to appear in ChatGPT answers?
Not in organic responses. Some assistants are testing sponsored placements, but the recommendations in normal answers are drawn from training data and retrieved web content, which is what this guide addresses.
How often do ChatGPT answers change?
Answers vary between sessions and change as models update and as retrieved sources change. That is why tracking should use a fixed prompt set, repeated monthly, and report trends rather than single results.
Does blocking GPTBot stop ChatGPT from mentioning us?
No. Blocking the crawler limits what it can read from your site, but the assistant can still describe you from other sources. Blocking usually makes descriptions less accurate, not less frequent.
What is LLM SEO?
LLM SEO, also called answer engine optimization, is the practice of increasing how often and how accurately large language model assistants mention and recommend your brand.
Written by
AI SEO Growth Agency
Senior SEO and AI search strategists who run growth programmes for SaaS, ecommerce and professional services brands. We publish what we learn from client work every week.
About the team
