To get your brand cited in ChatGPT, rank well in Bing and Google, publish answer-first pages with real statistics and sources, keep those pages fresh and earn mentions on the sites ChatGPT already cites, including Reddit. Then measure your mention rate across many runs, because a single ChatGPT answer tells you very little. The ten tactics below are ranked by how strong the evidence is, and each one comes with a way to check if it worked.
No tactic can guarantee a citation. But the research points clearly in a few directions, and you can test each one on your own prompts.
TL;DR
- Rank well in Bing and Google, because ChatGPT draws on search results.
- Write answer-first pages with real statistics and sources, and keep them fresh.
- Earn mentions on the sites ChatGPT already cites, including Reddit.
- Measure over many runs, because one answer tells you very little.
The 10 ways at a glance
| # | Tactic | Main evidence | Strength |
|---|---|---|---|
| 1 | Rank in Bing, not just Google | Seer: 87% of SearchGPT citations matched Bing top results | Strong |
| 2 | Measure mention rate across many runs | SparkToro and Gumshoe: same brand list in fewer than 1 in 100 runs | Strong |
| 3 | Earn branded mentions across the web | Ahrefs: 0.664 correlation with AI Overview visibility (75k brands) | Medium for ChatGPT |
| 4 | Add statistics, quotes and sources | GEO-bench: up to 40% higher visibility in tests | Medium |
| 5 | Keep pages fresh | Ahrefs: ChatGPT cites newer pages than Google organic | Medium |
| 6 | Show up in Reddit discussions | Semrush (325k prompts) and Peec: Reddit is a top cited domain | Medium |
| 7 | Get onto the pages ChatGPT already cites | Follows from how ChatGPT uses search results | Medium |
| 8 | Do not keyword stuff | GEO-bench: stuffing lowered visibility | Medium |
| 9 | Write answer-first openings | Best practice; no direct study | Low |
| 10 | Publish fair comparison pages | Best practice; no direct study | Low |
"Strong" means a large study with a direct link to AI citations. "Medium" means a lab test, a related engine or a correlation. "Low" means sound practice without a direct study.
AgenticSEO asks ChatGPT and Gemini your buyers' questions several times each week and shows your mention rate per prompt.
Check your AI visibility freeFree plan · No AI key to set up
1. Rank in Bing, not just Google
What to do: Make sure your key pages are indexed and ranking in Bing. Set up Bing Webmaster Tools, submit your sitemap and fix any crawl issues it reports.
Why: When ChatGPT searches the web, it relies on search results. Seer Interactive found that 87% of SearchGPT citations matched Bing top results. Many SEO teams focus only on Google and never check Bing.
Evidence strength: Strong.
How to check: Search your target prompts in Bing. Note where your pages rank. Compare with the pages ChatGPT cites for the same prompts. If ChatGPT cites pages that outrank you in Bing, improving your Bing position is a clear next step.
2. Measure mention rate across many runs
What to do: Ask each of your target prompts several times, not once. Record how many runs name or cite you. That share is your mention rate.
Why: ChatGPT answers change constantly. SparkToro and Gumshoe found that the same prompt gave the same brand list in fewer than 1 in 100 runs. One check can make you think you are winning or losing when you are not. Every other tactic on this list depends on measuring well.
Evidence strength: Strong.
How to check: Pick 15 to 20 buyer prompts. Ask each one 3 to 5 times in fresh chats. Record mention rate, competitors named and cited domains. Repeat weekly. Read more on why AI answers change and how mention rate works.
3. Earn branded mentions across the web
What to do: Get your brand named on other trusted sites: review sites, comparison articles, industry blogs, podcasts, newsletters and directories.
Why: Ahrefs studied 75,000 brands and found that branded web mentions correlate 0.664 with AI Overview visibility. That study is about Google AI Overviews, not ChatGPT. But it shows that AI systems lean toward brands the web talks about.
Evidence strength: Medium for ChatGPT. It is a correlation on a related engine.
How to check: Count new mentions each month. Watch whether your mention rate rises on prompts where those sites are cited.
4. Add statistics, quotes and sources
What to do: Put real, sourced numbers in your pages. Quote named experts or studies. Link to the original source.
Why: The GEO-bench paper (KDD 2024, arxiv 2311.09735) tested content changes on generative engines. Adding statistics, quotes and sources raised visibility by up to 40% in its tests.
Evidence strength: Medium. These were controlled tests, not live ChatGPT data.
How to check: Update a few pages with real data. Re-run their target prompts for two to four weeks. Compare mention rates with pages you did not change.
Never invent a statistic or a quote. Made-up facts hurt trust and can spread into AI answers.
5. Keep pages fresh
What to do: Review your most important pages on a schedule. Update facts, prices, screenshots and examples. Show the date you last updated the page.
Why: Ahrefs found that ChatGPT cites newer pages than Google organic results do. An old page may still rank in Google but lose out in ChatGPT.
Evidence strength: Medium.
How to check: Look at the publish and update dates of pages ChatGPT cites for your prompts. If they are newer than yours, a real update (not just a new date) is worth testing.
6. Show up in Reddit discussions
What to do: Find threads where people ask about your problem space. Answer honestly and helpfully. Mention your product only when it truly fits, and say you work there.
Why: Semrush (in a study of 325,000 prompts) and Peec both found that Reddit is one of the top cited domains in AI search.
Evidence strength: Medium.
How to check: Note how often Reddit appears in the cited domains for your prompts. Track whether threads that name you start to appear as sources.
Do not spam. Communities remove promotional posts, and a bad reputation there can follow you.
7. Get onto the pages ChatGPT already cites
What to do: For each gap prompt, list the domains ChatGPT cites. Then find honest ways to be included on those pages: pitch a review, ask to be added to a comparison, offer data or a guest post.
Why: If ChatGPT already trusts a page for a prompt, being named on that page puts you inside the answer's source set. This follows from how ChatGPT builds answers from search results.
Evidence strength: Medium. It is logical and tied to the Seer finding, but not tested on its own.
How to check: Re-run the prompt after you are added. Watch whether you are named when that domain is cited.
8. Do not keyword stuff
What to do: Write for readers. Use your main phrase where it fits naturally, in the title, the opening and a heading or two.
Why: The same GEO-bench paper that found gains from statistics and sources found that keyword stuffing lowered visibility.
Evidence strength: Medium.
How to check: Read your page out loud. If the keyword sounds forced, cut it. Then track the page's mention rate after the edit.
9. Write answer-first openings
What to do: Answer the question in the first two or three sentences. Then explain, compare and give detail.
Why: ChatGPT writes short answers and often draws on a short, clear passage. A page that gets to the point gives it something easy to use. We know of no direct study on this, so treat it as best practice.
Evidence strength: Low.
How to check: For each target page, ask: if someone read only the first paragraph, would they have the answer? Then compare mention rates before and after the rewrite.
10. Publish fair comparison pages
What to do: Create honest "X vs Y", "best tools for Z" and "alternatives to X" pages. Give each option real pros and limits, including your own product.
Why: Buyers ask ChatGPT comparison questions all the time. If you have a clear, fair page for that question, you give the AI a source that fits. Fairness matters, because a page that only praises you is less useful. This is best practice, not a tested finding. Our comparison page templates for SaaS can help you start.
Evidence strength: Low.
How to check: Track the comparison prompts the page targets. See whether ChatGPT names you more often, and whether it cites the page.
What does not seem to help?
Some popular ideas have little evidence behind them:
| Tactic | Finding |
|---|---|
| Schema markup | Ahrefs matched test: no clear effect on AI citations |
| llms.txt | SE Ranking (300,000 domains): no clear effect |
| Special AI files for Google | Google: no special files are needed for AI features |
These are not harmful, and schema can help other search features. Just do not count on them to get you cited.
Why is getting cited worth it?
AI traffic is small for most sites, but it can convert well. In a Seer Interactive B2B case, ChatGPT traffic converted at 16% vs 1.8% for organic. Ahrefs reported that 0.5% of sessions from AI gave 12.1% of sign-ups. These are single cases, so treat them as signs, not promises. Still, they explain why a citation can be worth more than its traffic suggests.
A simple weekly routine
| Day | Task |
|---|---|
| Monday | Re-run your prompts and record mention rates |
| Tuesday | Pick one or two gap prompts and list their cited domains |
| Wednesday | Update or write one answer-first page |
| Thursday | Pitch one site that ChatGPT already cites |
| Friday | Answer one or two relevant community questions |
If you want the measuring done for you, AgenticSEO asks each prompt several times in ChatGPT and Gemini, shows your mention rate, competitor share and cited domains, and turns gaps into tickets. It can then write, QA-check and publish the fix. You can start free with 10 tracked prompts.
For the full picture of how AI search works, read what generative engine optimization is. SaaS teams can also read how SaaS companies get recommended by AI.
Ready to turn gaps into pages? AgenticSEO writes each one, checks every claim against your approved facts, and holds it in Ready for your approval unless you turn on auto-publish. Compare plans
Frequently asked questions
Get cited in ChatGPT and Gemini. Every week.
Agents find the prompts you lose, write fact-checked pages, publish them and measure again. The AI is included.