# How to get recommended by ChatGPT (and the other AI assistants)

How ChatGPT, Gemini, Claude, Perplexity and Muse pick which products to name, what actually moves it, and what is sold but does not work.

By Dasha. Updated November 1 2026. https://heardofyou.io/get-recommended-by-chatgpt

To get recommended by ChatGPT, you have to be in the sources it reads when a buyer asks. Those are mostly other people's pages: "best X" lists, review sites, Reddit threads and YouTube videos. The AI searches, reads what comes back, and names the products it keeps seeing.

So the work has three parts. Show up in the AI's search results. Get genuinely mentioned on the pages it cites for your category. Make your own site easy to quote. Then measure, because answers move.

That is the whole method. The rest of this page is the evidence, the traps, and what we do about it.

## Why it matters now

51 percent of B2B software buyers now start their research in an AI chatbot more often than in Google. In the same G2 research, 69 percent said a chatbot led them to a different vendor than planned. One in three bought from a vendor they had never heard of. The shortlist inside the answer is where that happens.

**51% : of B2B software buyers now start research in an AI chatbot more often than Google : G2 Answer Economy report, March 2026**
## How an AI decides what to name

1. **It searches first.** For most buying questions, ChatGPT, Gemini and Perplexity search the web before answering. For software questions, 76 percent of ChatGPT's searches add the current year ("best invoicing app 2026"). Missing from those results, you are mentioned about 2.8 percent of the time. With your site cited and you in branded searches, that rises to 91.4 percent.
2. **It trusts what others say.** For US software questions, 72.7 percent of the sources AI engines cited were earned media, not the brand's own pages. Across 207 million citations, more than 84 percent were earned or third-party. 0.3 percent were paid.
3. **It breaks ties on reputation.** In a 2026 experiment with identical specs, the well-known brand won 100 percent of the time. A rating edge of under 0.1 stars for the challenger flipped it. Reviews are the tie-breaker a small product can actually earn.
4. **It quotes what is quotable.** In the original GEO study, pages with specific numbers, quotes and named sources were 30 to 40 percent more visible. Keyword stuffing did nothing.
5. **It changes its mind.** Run the same prompt twice and the same brand list comes back less than one time in a hundred. Category leaders still show up most of the time. The margins move. That is why we report how often you are named, never a rank.

## What works

- **Get onto the lists the AI cites for your category.** Not every list. The three to ten that appear in the citations for your buyer's questions. Pitch them with a real reason. Follow up twice.
- **Get real reviews on the review site that shows up most.** For B2B that is usually G2 or Capterra. For apps, the App Store or Product Hunt. For Perplexity especially, G2.
- **Answer real Reddit threads as yourself.** Disclosed, useful, no link unless it helps. Reddit is in every engine's top five.
- **Put the facts in your homepage's first paragraph.** What it is, for whom, how much, what it does not do. Most AI referrals land on the homepage.
- **Write one honest comparison page and one alternatives page.** "Comparison" is one of the most common words in ChatGPT's own search queries.
- **Let the crawlers in.** Block Claude's crawler and you get almost no Claude citations. Check robots.txt. Make sure your words are in the HTML, not loaded afterwards.
- **Rank in ordinary search for the questions the AI searches.** It still feeds the machine.

## What does not work (but is widely sold)

| Sold as | What the evidence says |
|---|---|
| An llms.txt file | 97 percent of the llms.txt files Ahrefs tracked got zero requests. Harmless, cheap, not a lever. We ship one on our own site and say so. |
| Schema markup alone | Ahrefs tracked 1,885 pages that added schema: no statistically significant change in ChatGPT citations. Fine for other reasons, not for this one. |
| Directory blasts | ChatGPT and Gemini rarely cite directories (9,978 answers). |
| "AI-optimised" rewrites | Most AI-specific content tricks were largely ineffective in the NeurIPS 2025 benchmark. Being in the sources beat any rewrite. |
| Guaranteed #1 in ChatGPT in 14 days | Answers do not repeat run to run. Nobody can guarantee this, and anyone who says so is the red flag. |
| Fake reviews, persona accounts, hidden text | Works in lab conditions, counts as manipulation, risks your name. We do not do it. |

## What we would do

**Free check.** Your five most important buyer questions, three runs each, on ChatGPT, Gemini, Claude, Perplexity and Muse: 75 answers. Per question, you see who was named instead of you and which sources were cited. By email within 2 working days. A person checks every report.

**Audit, $149** ($99 for the founding 20). 25 questions, same method. Plus a prescription: the exact lists, review sites and threads to get into for your category, ranked by how often the AIs cited them. Credited toward the project within 30 days.

**Project, $799.** We pitch 25 named lists and sites, with two follow-ups each. We write three pages for you: your story page, one honest comparison, one alternatives page. We set up review profiles on the two sites AI cites most in your category. You get a founder-led Reddit plan. We re-ask the same questions at day 30, 60 and 90.

**Keep growing, $349 a month**, 3-month minimum, offered after the day-30 re-check. One new placement a month plus a monthly re-check.

Every deliverable ships within 90 days or your money back. The day-60 target: named in at least your baseline plus 3 of the 20 agreed answers. Never fewer than 3, never more than 7. Miss it, and we keep working at no charge until day 120. We never promise what an AI will say.

Example: "Ledgerly" (a made-up invoicing app for freelancers). Free check: named in 12 of 60 answers. ChatGPT 5 of 15, Gemini 3 of 15, Claude 3 of 15, Perplexity 1 of 15. Named instead: FreshBooks, Wave, Bonsai. Sources cited: G2, a Zapier "best of" list, two Reddit threads. The to-do list writes itself from that last line.

Paying is simple at launch. You fill in a short request form. I confirm fit within 2 working days and send an invoice. No card on file, no subscription you forget about.

## Questions

**Does submitting to directories help AI visibility?**

Rarely. In a study of 9,978 AI answers, ChatGPT and Gemini seldom cited directories. Directories can nudge your domain rating for ordinary search. They are not where the AI gets its shortlist, so we never sell directory submission.

**Should I just buy ChatGPT ads instead?**

You can, from $25 a day. OpenAI says the ads do not change the organic answer. Ads stop the day you stop paying. A place on a list the AI cites keeps working. They are not substitutes. If you can afford both, do both.

**Can anyone guarantee ChatGPT will recommend me?**

No. Anyone who does is selling something else. Answers change run to run, and models update on their own schedules. What can be promised is the work (named placements, pages, profiles) and honest measurement before and after. That is what we promise.


## Sources

- G2 research, The Answer Economy: 51% of B2B software buyers start in an AI chatbot more often than Google (https://company.g2.com/news/g2-research-the-answer-economy)
- Tannenbaum, fitted stage model: 2.8% vs 91.4% mention rate (2026, pre-print) (https://arxiv.org/abs/2609.23162)
- Chen et al., University of Toronto: 72.7% of AI sources are earned media (https://arxiv.org/html/2509.08919)
- Muck Rack Generative Pulse: 84%+ earned, 0.3% paid; blocking Claude's crawler means almost no Claude citations (https://www.globenewswire.com/news-release/2026/10/28/3390134/0/en/generative-pulse-earned-media-still-drives-more-than-84-of-ai-citations.html)
- Aggarwal et al., GEO (KDD 2024): quotes, statistics and citations raised visibility 30 to 40%; keyword stuffing did not (https://arxiv.org/html/2311.09735v2)
- C-SEO Bench (NeurIPS 2025): most AI-specific content tricks were largely ineffective (https://ar5iv.labs.arxiv.org/html/2506.11097)
- Ahrefs: we tracked 1,885 pages adding schema; AI citations barely moved (https://ahrefs.com/blog/schema-ai-citations/)
- Ahrefs llms.txt data (97% of files got zero requests), via YourWebTeam (https://yourwebteam.io/ai-seo-scams-small-business-2026/)
- GetIntel, 9,978 AI answers: ChatGPT and Gemini rarely cite directories (https://getintel.ai/blog/best-saas-directories-2026/)
- Chu and Hou: brands win ties, a rating edge under 0.1 stars flips it (2026, pre-print) (https://arxiv.org/abs/2606.17443)
- SparkToro: same brand list less than 1 time in 100 (https://sparktoro.com/blog/new-research-ais-are-highly-inconsistent-when-recommending-brands-or-products-marketers-should-take-care-when-tracking-ai-visibility/)
- OpenAI: testing ads in ChatGPT (ads do not change the answer) (https://openai.com/index/testing-ads-in-chatgpt/)
- StatWharf: 76% of ChatGPT's software searches add a year (https://statwharf.com/research/chatgpt-query-fan-out/)