AI Visibility Research: 10 Lessons From 103,427 AI Answers
Lesson 6: "Best tools" listicles win the most competitor citations
We sorted the 24 most cited competitor URLs by page type. Together they earned 22,884 citations in the window.
| Page type | URLs | Citations | Share |
|---|---|---|---|
| "Best tools" listicles | 8 | 6,962 | About 30% |
| Product, feature and home pages | 5 | 6,941 | About 30% |
| How-to guides | 6 | 4,936 | About 22% |
| Original research pages | 2 | 1,832 | About 8% |
| Free tools | 2 | 1,360 | About 6% |
| Help center articles | 1 | 853 | About 4% |
The single most cited competitor URL was a listicle of Google AI Overviews trackers on llmpulse.ai, with 1,988 citations.
A how-to guide on tracking Google AI Overviews from otterly.ai came second with 1,792. Both answer a narrow buyer question in a format you can scan.
Listicles win because a "best tool for" prompt is a request for a shortlist. A page that already has one, with criteria, is easy for an engine to reuse.
Lesson 7: ChatGPT often cites product pages and free tools
Product pages earned almost as many citations as listicles. Profound’s citations feature page drew 1,690 and the Otterly homepage drew 1,669.
ChatGPT leaned toward product pages most. Its most cited competitor URLs included Ahrefs Brand Radar, the Ahrefs AI Overviews tracker and an Otterly feature page.
Free tools got cited too. The Ahrefs AI visibility checker earned 948 citations and the Semrush AI search visibility checker earned 412.
In ChatGPT, our second most cited URL was a feature page for citation monitoring.
So write feature pages in plain text: what the product does, who it’s for and how it works. An engine can’t quote a feature it can’t read.
Lesson 8: Original research earns citations
Two research pages made the top 24: a Semrush study of AI Overviews with 1,271 citations and the Ahrefs AI visibility index with 561.
Academic sources show up as well. arxiv.org held a 3.2% citation share among third-party domains in our data.
Peer-reviewed work points the same way. The GEO paper, accepted at KDD 2024, found that optimization can raise visibility in generative engine responses by up to 40%.
The methods that helped in that study include adding statistics, quotations and cited sources. Original data hands engines a fact no other page has.
Lesson 9: A few pages carry most citations, so own a niche
AI engines cited ansvisor.com pages 19.7k times in the window, and a small group of pages earned most of them.
One page, our list of top GitHub repositories for AEO and GEO, earned 7,007 citations. That’s more than the next five pages combined.
Topic data shows the same pattern. A brand’s visibility on a niche topic it owns can be more than ten times its visibility on broad topics.
Broad definition prompts such as "What is GEO?" and "What is AEO?" are much harder to own.
Engines reward the source that owns a specific question. Find the questions where you know something others don’t, and build the definitive page for each one.
Lesson 10: AI visibility changes from month to month
Between two 30-day windows, a brand’s mentions and citations can more than double.
Short windows move even more. A single day’s visibility score can swing well away from the 30-day figure.
We can’t pin these shifts on one cause. Content changes, competitor updates and model or retrieval changes all play a part.
So a one-off check misleads. Run the same prompts every week and the trend shows up.
How can a company improve its visibility in AI-generated answers?
Use the lessons above as a checklist. Each step maps to one finding from the data.
Pick specific questions you can own and build one strong page per question (Lesson 9).
Put the direct answer and the facts that matter at the top of each section.
Add numbers and quotes with named sources that engines can lift into an answer (Lesson 8).
Publish a shortlist with stated criteria for the "best tool for" questions in your category (Lesson 6).
Describe features, use cases and limits in plain text on your product pages (Lesson 7).
Show up with honest, useful content on the third-party sites engines cite most (Lesson 5).
Name your brand next to your facts, so a citation also counts as a mention (Lesson 2).
Placement matters too. The "Lost in the Middle" study found that language models use information at the start or end of long inputs best.
The effort pays off because answers absorb clicks. A Pew Research Center study found users clicked a result in 8% of visits with an AI summary, versus 15% without.
For more tactics, see our guide to the best ways to increase AI citations.
How to benchmark your brand visibility in AI answers
You can repeat this research for your own category. Use the template below and keep it identical on every run.
Prompt set: 50 to 200 prompts across definitions, how-to, "best tool for" and comparison questions.
Platforms: ChatGPT, Google AI Overviews, Google AI Mode, Gemini, Microsoft Copilot and Claude, tracked separately.
Brands: you plus 5 to 15 competitors, with every name variant listed.
Metrics: mentions, citations, prompt coverage, share of voice and links per answer.
Sources: your cited URLs, competitor URLs grouped by page type, and the top third-party domains.
Cadence: weekly runs, with a monthly summary of changes.
Our guide on how to measure AI share of voice covers the calculations step by step.
How Ansvisor helps you run your own AI visibility research
Ansvisor runs your prompt set on a schedule and stores every answer, along with each brand named and each URL cited. The numbers in this report come from that workflow. Claude tracking is available on Enterprise plans.
Answer Engine Insights shows the brand leaderboard and share of voice by platform. Competitor benchmarking puts your mentions and citations next to each competitor’s on the same prompts.
Use the cloud version or self-host it, and benchmark your own category this week.
FAQ
How can a company improve its visibility in AI-generated answers?
Build one strong page for each specific question you can own, and answer it in the first sentences. Add numbers and quotes with named sources. Earn useful mentions on sites engines already cite, then track the same prompts weekly.
How do AI model updates affect visibility?
Model and retrieval updates can change which sources an engine uses and which brands it names. You rarely see the update itself, only what it does to answers. A fixed prompt set tracked over time is the most reliable way to spot those shifts.
Can media coverage improve visibility in AI-generated answers?
It can. In our tracking data, third-party domains such as Reddit, YouTube, GitHub, LinkedIn and Search Engine Land were among the most cited sources. Coverage on sites that engines already cite gives them another source that names your brand.
What is the difference between an AI mention and an AI citation?
A mention is your brand name in the answer text. A citation is a link to one of your URLs shown as a source. One can rise without the other, so measure both for every prompt and platform.
How do you benchmark brand visibility in AI answers?
Run a fixed set of buyer prompts on each AI platform on a schedule. Record mentions, citations, prompt coverage and share of voice for you and your competitors. Compare the same metrics every week and summarize changes monthly.
