How to get your product recommended by ChatGPT

A growing share of developers now ask an AI model which tool to use before they ever open Google. “What’s the best way to add auth to a Next.js app?” “Datadog alternatives for a small team?” If ChatGPT answers with three products and yours isn’t one of them, you lost the deal without knowing it was there.

There’s no ad slot to buy and no magic tag to add. But where these answers come from is fairly well understood, and you can work on every one of those sources.

Where AI answers come from

Every AI answer is built from one or both of two places:

1. Training data. What the model absorbed about your product during training: how often it’s discussed, in what context, and whether people like it. This changes slowly, only when a new model is released.
2. Live retrieval. When ChatGPT, Perplexity, Gemini or Google AI Overviews search the web for a question, they read a handful of pages and summarize them. This can change within weeks.

For retrieval, each engine uses a search index. Google AI Overviews and AI Mode use Google’s. ChatGPT search uses its own crawler (OAI-SearchBot) and third-party search providers. Perplexity runs its own crawler and index. In practice, if you don’t rank in normal search for a question, you’re unlikely to be retrieved for it either.

What gets retrieved and quoted tends to be:

• Discussions. Reddit threads, Hacker News, Stack Overflow, GitHub issues. Google and OpenAI both have data agreements with Reddit, and Reddit threads show up constantly as cited sources for “which tool” questions.
• Comparison and “best of” articles, including ones written by your competitors.
• Documentation and clear product pages that state plainly what the product does, who it’s for, and what it costs.
• Review and directory sites where a category is listed.

The steps that actually move it

1. Find out where you stand

Write down 20 to 50 questions your buyers actually ask, in their words: “best open source feature flag tool”, “LaunchDarkly alternative that’s cheaper”, “how to do feature flags in Go”. Ask them in ChatGPT, Perplexity, Gemini and Google, and write down who gets named, in what order, and which sources get cited. Do it more than once, because answers vary between runs. How to track AI visibility covers doing this properly.

The cited sources are your to-do list. If the same three Reddit threads and one comparison article keep appearing, those are the pages that decide the answer.

2. Be in the threads the engines cite

If a Reddit thread is cited for “best feature flag tool for Go” and you’re not in it, a genuinely useful comment from someone who has used your tool is the most direct lever you have. It adds your product to the exact text the engine is summarizing. The same goes for new threads: comparison threads that start ranking this month are next quarter’s citations.

This is where most of our work goes, and the method is in Reddit marketing for developer tools. The rule is the same as always: answer first, be specific, name competitors honestly, disclose affiliation.

3. Make your own site easy to quote

AI engines extract sentences, not pages. Help them:

• Say what the product is in one plain sentence near the top of the homepage: “X is an open source feature flag service for Go and TypeScript teams.” Not “Ship with confidence.”
• Publish pages for the questions buyers ask: comparisons (“X vs Y”), alternatives, use cases, pricing. Put the direct answer in the first paragraph.
• Keep pricing, limits and integrations in crawlable HTML, not only in images or behind a login.
• Add structured data (Organization, SoftwareApplication, FAQPage) so the facts are unambiguous.
• Make sure AI crawlers are allowed. Check robots.txt for GPTBot, OAI-SearchBot, ClaudeBot and PerplexityBot, and check your CDN. Some block AI crawlers by default.

4. Get mentioned in places you don’t own

Engines trust consensus. One page on your site saying you’re the best is weak. Five independent sources mentioning you for the same use case is strong. Useful sources:

• Relevant Reddit and Hacker News discussions
• Comparison articles and “best X tools” lists (reach out to the authors with accurate info)
• Directories and review sites for your category
• Integration pages on partners’ sites
• Answers on Stack Overflow and GitHub discussions where your tool is the genuine fix

5. Fix how you’re described

Sometimes you’re mentioned, but wrongly: an old price, a missing feature, “expensive” when you’re not. Find the source the engine is paraphrasing (Perplexity and ChatGPT search show citations) and fix it there. Update your own page, reply in the thread with a correction, or ask the article’s author to update.

What not to do

• Don’t spam AI-generated pages. Hundreds of thin “best X” pages on your own domain rarely get cited, and they can hurt your normal rankings.
• Don’t fake reviews or discussions. Coordinated fake Reddit activity gets removed, and removed content can’t be cited.
• Don’t chase a single prompt. Answers vary by wording, user and day. Track a broad set of questions and look at the trend.

How long it takes

Retrieval-based answers can pick up a new thread or page within weeks once it ranks. Training-data effects lag by months. In our experience the curve usually bends around month three of consistent work. Brands that stop after one month rarely see it.

If you’d rather have someone run this, it’s exactly what answer engine optimization for developer tools looks like in practice.