How to get your product recommended by ChatGPT, Perplexity and other AI search

More people now ask an AI assistant "what's the best [product] for [need]?" instead of scrolling search results. If your product isn't in those answers, you're invisible to a growing share of buyers. Nobody outside the model providers knows the exact ranking rules, and they change. But the patterns that make a brand more likely to be mentioned are consistent enough to act on.

How these assistants find answers

Roughly, two things feed the answer:

1. What the model learned in training: text from across the web, including reviews, articles, forums and documentation.
2. What it looks up at the moment you ask: for assistants with web search, live results from search indexes, which they read and summarise.

You can influence both. For the first, be present and consistently described across the web. For the second, rank for and be easy to read on the queries people ask.

What tends to help

Clear, consistent descriptions of what you are

Say plainly, on your homepage and in your docs: who it's for, what it does, and what category it's in. Use the same phrases everywhere (website, directories, social profiles). Ambiguity makes it hard for a model to categorise you.

Third-party mentions

Assistants lean on what other sources say about you. Useful places include:

• Review sites and directories relevant to your category.
• "Best of" and comparison articles by independent publishers.
• Reddit, forums and Q&A sites where people discuss tools.
• Newsletters, podcasts and YouTube reviews.
• Wikipedia or Wikidata entries, where you legitimately qualify.

Earn these; don't fake them. Astroturfing gets detected and can backfire.

Pages that answer the questions people ask

Create content that directly addresses "best X for Y", "X vs Y", "how to do Z" and "alternatives to [competitor]". Answer the question in the first paragraph, then support it. Be fair in comparisons; balanced, factual pages are more likely to be cited.

Structured, readable pages

• Descriptive headings, short paragraphs and lists.
• A FAQ with real questions and direct answers.
• Pricing, features and limits stated clearly in text, not hidden in images.
• Schema markup (Product, Organization, FAQ) where appropriate.
• Fast, accessible pages that crawlers can read without JavaScript tricks.

Being crawlable

Check your robots.txt isn't blocking legitimate AI and search crawlers you want to be found by. Decide deliberately which bots to allow. A proposed llms.txt file may help some tools; see llms.txt explained.

Traditional SEO still matters

Assistants with web search often pull from top-ranking pages. Good SEO (first 100 days) feeds AI visibility.

What doesn't work

• Hidden text or prompts aimed at models. Platforms treat this as manipulation.
• Paying for fake reviews.
• Keyword-stuffed pages.
• Assuming one page will "rank" for an AI; there is no single ranking.

How to check where you stand

Once a month, ask the major assistants the questions a buyer would ask ("best [category] for [use case]", "[competitor] alternatives", "is [your brand] good?"). Record whether you're mentioned, how you're described, and what sources are cited. Compare over time. Answers vary between runs, so ask a few times and look at patterns, not single results.

Fix what you find

• Not mentioned: build third-party presence and comparison content.
• Described wrongly: update your site and profiles, and correct facts where sources are wrong.
• Mentioned but negatively: address the underlying issue and reviews.

Realistic expectations

This is a slow game, measured in months. The good news is that the work overlaps with what already helps customers and search: clear positioning, honest content and real reputation.