← Blog · 2026-10-17 · 4 min read
How to Get Mentioned by ChatGPT - 9 Fixes That Actually Move AI Recommendations
There's no form you can fill in to get ChatGPT to recommend your product. AI assistants recommend what the web already says about you: on your own site, and far more importantly, on the pages they trust. The good news is that most of those pages are within your reach.
Here are nine fixes, ordered roughly by impact for a small SaaS or digital product.
1. Find out where you stand first
Before you fix anything, measure. Ask the 5 to 10 questions your buyers would ask ("best [category] for [audience]", "alternatives to [competitor]") across ChatGPT, Perplexity, Gemini and Claude. Record whether you're mentioned, your position, who's named instead, and which sources are cited. (Here's the full method, or run a free automated scan.)
The cited sources are your roadmap. Everything below is about getting onto those sources.
2. Make sure AI crawlers can actually read your site
This sounds basic, but it's a surprisingly common reason products get skipped. Check your robots.txt for blocks on:
GPTBotandOAI-SearchBot(OpenAI)PerplexityBotClaudeBot(Anthropic)Google-Extended(Gemini training)
Also check your CDN. Some hosts and Cloudflare offer one-click "block AI bots" settings. If you want to be recommended, don't block the assistants doing the recommending.
3. Say what you are in one plain sentence
Assistants pull short, quotable descriptions. Your homepage H1 and meta description should say, in plain words:
[Product] is a [category] for [audience] that [main result].
"The future of work, reimagined" gives an AI nothing to quote. "Acme is an invoicing tool for freelancers that gets you paid in half the time" does. Repeat the same sentence on your directory listings and social bios so the web agrees on what you are.
4. Put pricing, features and FAQs in plain HTML
If your pricing only appears inside a JavaScript widget, or your features sit in images, many crawlers can't see them. Use real HTML text for:
- pricing tiers and what's included,
- a feature list,
- an FAQ that answers real buyer questions word for word ("Does Acme work with Stripe?").
Add SoftwareApplication, Organization and FAQPage schema. An llms.txt file (a short Markdown summary of your product at /llms.txt) costs you five minutes, so it's worth adding.
5. Get onto the review sites AI cites
In software categories, G2 and Capterra pages show up constantly as AI sources. A claimed profile with even 5 to 10 genuine reviews is often the single biggest lever for a new product. Ask your happiest customers directly and send them the review link. Don't offer incentives that break the review site's rules.
6. Show up in the Reddit threads that get cited
Perplexity and ChatGPT search frequently cite Reddit threads like "What do you use for X?". Find the ones that rank for your category and answer them honestly, as the maker:
"I built [product], so I'm biased. It's good for X but not ideal if you need Y. For Y, [competitor] is better."
Honest, specific answers get upvoted and survive moderation. Drive-by self-promotion gets removed and can hurt you.
7. Publish honest comparison pages
When buyers ask "Alternatives to [Big Competitor]", assistants love pages that answer that exact question. Publish:
/alternatives/[competitor]: who should pick you instead, and who shouldn't./vs/[competitor]: a fair feature and price table.
Be accurate and date your price checks. Exaggerated comparison pages erode trust with readers and with the third-party writers you want citing you.
8. Get into the roundups
"Best [category] tools in 2026" articles are among the most cited pages in AI answers. Search Google for those phrases and email the authors of the top 10 results with a short, specific pitch. Include one sentence on what you do, who it's for, and what makes you different, plus a free account. Many roundups are updated regularly, and authors need new tools to add.
9. Measure every week with the same questions
AI answers are non-deterministic, so one check tells you little. Pick a fixed set of questions and track:
- mention rate (the % of answers that name you),
- average position,
- top rival's mention rate,
- which sources are cited.
Then ship one or two fixes a week and watch the trend. Live-search assistants (Perplexity, ChatGPT search) tend to respond in days to weeks. Training-data-based answers (Claude without search, older model snapshots) move over months.
A realistic timeline
| Week | Focus |
|---|---|
| 1 | Baseline scan, robots.txt check, homepage sentence, HTML pricing and FAQ, llms.txt |
| 2 | Claim G2 and Capterra and ask for reviews; list on 10 relevant directories |
| 3 | Answer 5 relevant Reddit threads honestly; publish 1 comparison page |
| 4 | Pitch 10 roundup authors; publish a 2nd comparison page; re-measure |
Track it automatically
AskRankly runs your buyer questions across ChatGPT, Perplexity, Gemini and Claude and shows your mention rate, rank, rivals and cited sources, with 3 to 5 fixes based on your actual data. The first scan is free, and weekly tracking starts at $19 a month.
See if ChatGPT recommends your product
Free scan across ChatGPT, Perplexity, Gemini and Claude in about 60 seconds.
Run my free scan