Guides · Updated November 13, 2026

How to rank in ChatGPT: get your product recommended

There's no single ranking in ChatGPT. For each question a buyer asks, it names a few products, and a different few the next time. “Ranking” means being named in most answers to the questions your buyers ask. Here's how we work on that for our own three products, with our real numbers, and what the evidence says works.

Where our products stand

We build Askadia, and we run it on our own products: PodSeeker (a podcast database for PR teams), SocialPhotos (a customer photo gallery for Shopify stores) and Askadia itself. Their free checkups, 10 buying questions each, asked once:

Three very different starting points: a product ChatGPT already knows, one it confuses with another category, and a new one it barely knows. We track each one's questions on a schedule: in the week to Oct 29, ChatGPT named PodSeeker in most answers to 18 of its 25 questions, and Askadia in 1 of 13; in the week to Nov 13, PodSeeker in 23 of 25 and Askadia still in 1 of 13. We write down what we change, and report what happens in our log below.

Eight steps

1. Pick the questions your buyers ask

You rank per question, so start with the questions. Write them the way your buyer would type them into ChatGPT, with their situation in it (“for a small PR agency”, “for a Shopify store”), not your marketing words or a search keyword. Ten to twenty is plenty.

On ours: our first PodSeeker questions included our own homepage phrases and a budget picked to match our price. ChatGPT named us, but we'd built the win into the question. We replaced them with the words our customers used in reviews and support emails.

2. Measure before you change anything

Ask each question once a day for a week before you trust a result. One answer is a sample: across thousands of runs, ChatGPT gave the same list of brands less than 1% of the time (SparkToro). Count a question as won when most of its answers name you, and read the count across all your questions.

On ours: in PodSeeker's first days of daily answers (Oct 25–29, 2026; 24 questions with 3 or 4 answers each), ChatGPT named PodSeeker in every answer to 16 questions and in none to 2. On the other 6 it changed its mind: on one question it named us once in 4 answers, on another 3 times in 4. Judged on a single answer, any of those 6 could have gone either way.

3. Let ChatGPT read you

Check that your robots.txt allows OAI-SearchBot (the crawler behind ChatGPT search; blocking GPTBot only opts you out of training), that your CDN doesn't block AI crawlers, and that your key pages work without JavaScript. The checklist is in our guide to why ChatGPT leaves products out.

4. Say plainly what you are and who it's for

When ChatGPT searches, it shortlists products, then checks their own websites (Minuttia, 18,084 software questions). Your homepage, pricing and feature pages should say, in your buyer's words: what the product is, who it's for, what it costs, what it works with, and how it differs.

On ours: ChatGPT read some of PodSeeker's pages and still named someone else. We rewrote them to answer the buyer's question directly.

5. Get on the pages it reads

For software questions, the pages ChatGPT cites most are other people's “best X” lists: 43.8% of cited pages (Ahrefs). In Ahrefs' own test, 94% of an established product's new mentions came from other people's pages (Ahrefs). List the pages cited across your questions, and ask each author to include you, honestly, where you fit.

On ours: one guide was cited in PodSeeker's answers again and again, and it didn't list us. We wrote to its author.

6. Fix your pages elsewhere

Your app-store listing, marketplace page or directory profile may be what ChatGPT reads about you, more than your site. Treat its first lines like your homepage's.

On ours: SocialPhotos' most-cited page was its Shopify App Store listing: counting it as ours, SocialPhotos was cited in 9 of 16 questions, not 5. Next: reword the listing's first line, and list it in the Shopify App Store category its buyers browse.

7. Own a narrow question

When products look alike, ChatGPT names the one everyone knows; a clear, concrete difference breaks that (Chu & Hou, 2026). Pick the buyer and need you serve best, and make that difference plain everywhere ChatGPT reads.

On ours: SocialPhotos' questions said “customer photos”, which ChatGPT read as reviews apps, so it recommended Loox and Judge.me. Its buyers say “customer photo gallery”, so that's the word we're moving to: in its questions first, then the listing and the site. Askadia's is “AI visibility tracker for software products”: not brand monitoring for marketing teams.

8. Change one thing, then wait

Answers that search the web pick up a changed page within days to weeks, and some never do; answers from memory change only with a new model. Write down what you changed and when, keep asking every day, and judge it after a few weeks.

What not to do

Listicles that rank yourself first, planted Reddit posts, bought links, fake reviews and hidden instructions for AI range from useless to harmful, and llms.txt and schema markup show no measured effect. The evidence for each is in what works, and what doesn't.

Questions founders ask

How long does it take to rank in ChatGPT?

For answers that search the web, a changed page can show up within days to a few weeks; studies put the typical lag at about one to two weeks, and some changes never show. Answers from ChatGPT's memory change only when a new model is trained. Judge any change over several weeks of daily answers.

Can you pay to rank in ChatGPT?

Not in the answers. ChatGPT shows ads to some logged-in users in the US, labeled and kept apart from the answer; OpenAI says ads don't influence what the answer recommends. Paid placements on other sites' lists exist, but there's no independent evidence they work, and undisclosed ones break the rules.

Does ranking on Google help you rank in ChatGPT?

Some, not much. Only about 8% of the pages ChatGPT cites are in Google's or Bing's top 10 for the question (Ahrefs): it runs its own searches and reads other pages. Being named on the pages it reads matters more than your own rankings. Google's own AI Overviews and AI Mode are another matter: they read Google's index (ChatGPT vs Google AI Mode).

How do I check whether ChatGPT recommends my product?

Ask your buyers' questions in the ChatGPT app, logged out, several times each, and count how often you're named and who's named instead. Askadia's free checkup does it for 10 questions in about 2 minutes, on Google's AI Overviews and AI Mode too, each counted on its own; tracking asks them on a schedule, on the assistants you choose.

Our log

What we changed on our own products, and what happened. We add to it as results come in.

• Oct 25–28, PodSeeker: replaced questions written in our own words; asked every question daily for a baseline week; rewrote pages ChatGPT read without naming us; wrote to the author of a guide cited again and again. Results, two weeks later: in the week to Nov 13, ChatGPT named PodSeeker in most answers to 23 of its 25 questions (18 in the week to Oct 29), made it the top pick on 16 (12 the week before) and cited its site on all 25. The clearest single change: “How do I find podcasts that are actively booking guests for my clients?” went from 0 of 1 answers naming us to 4 of 6 in the days after we rewrote the blog post ChatGPT read for it (Oct 27). A caveat on the weekly numbers: since Nov 1 each question is asked every few days, so a week holds 2 or 3 answers per question and one can flip on a single answer; we can't yet say how much of the rise is our changes and how much is ChatGPT changing its mind. The guide's author hasn't replied yet; we followed up on Nov 9.
• Oct 28–29, SocialPhotos: counted its Shopify App Store listing as its own page (cited in 9 of 16 questions, not 5); added questions in the buyers' words, “customer photo gallery”; tracked the two galleries that do the same job. Results: in a few weeks.
• Oct 29, Askadia: a baseline of 13 buying questions, asked daily (mentioned in 1 of 10 at the checkup, none of our pages cited); this guide and its companion. Results, two weeks later: nothing moved. In the week to Nov 13, ChatGPT named Askadia in 1 of 13 questions, never as the top pick, and cited none of our pages. Publishing guides on our own site didn't change that on its own, which matches step 5: the pages ChatGPT reads for these questions are other people's lists.
• Oct 31, PodSeeker: started asking its 25 questions on Google's AI Overviews and AI Mode too, every day. On the first day, one answer each, AI Overviews named it on 15 and AI Mode on 7, against ChatGPT's 19 that week (what we're seeing). Results: in the week to Nov 13, with 3 answers a question, AI Overviews named PodSeeker on 19 of 25 questions and AI Mode on 12, against ChatGPT's 23.
• Nov 12–13, Askadia: published our own category ranking (40 buying questions about AI visibility tools, on all three assistants), then wrote to the authors of six lists the assistants cite for those questions, none of which named us. Results: from late November.

That's how we ran PodSeeker's baseline week in the log above: our agent checked the questions, asked them daily and read every answer.