Competitor analysis for AI citations means asking AI engines the questions your buyers ask and recording who they recommend and which sources they cite when they don’t name you. Lock a set of 20 to 50 prompts, run each one several times on every engine that matters to you, and log every brand named and every domain cited. That gives you each brand’s appearance rate (share of answers naming it), citation rate (share linking to its site) and share of AI voice (its share of brand mentions across the competitors you track). The most useful output, though, is the list of pages and third-party sources cited instead of you: it tells you what to fix, publish or earn.

This page is the competitive method. The measurement basics it builds on (sample sizes, margins of error, what to log, free Search Console and Bing data) are in How to Measure AI Citations. Tool prices and features are in Best AEO Tools in 2026.

Why your AI competitors aren’t always your Google competitors

The brands that outrank you on Google and the brands an AI engine recommends overlap, but they aren’t the same list, for three reasons.

Engines cite pages that don’t rank for the query. In Ahrefs’ March 2026 study of 4 million URLs cited in AI Overviews, only 37.9% also ranked in the top 10 for the same query. Google’s AI features run extra “fan-out” searches, so a competitor can win with a page that ranks for a related question.

Each engine leans on different sources. In June 2025 data, only 7 of the 50 most-mentioned websites were on the lists for all three of AI Overviews, ChatGPT and Perplexity (Ahrefs). In Yext’s study of 6.8 million citations for location-based questions in four industries (July–August 2025), Gemini drew 52.15% of its citations from brand-owned websites while OpenAI drew 48.73% from listings, mostly Google’s.

Third-party coverage carries a lot of weight. Chen, Wang, Chen and Koudas (arXiv:2509.08919, September 2025) found AI search showed “a systematic and overwhelming bias towards Earned media” (third-party, authoritative sources) over brand-owned and social content. The competitor that wins AI answers is often the one review sites, comparison articles and forums talk about most.

So don’t assume your competitor list; discover it from the answers, engine by engine. More on what each engine cites is in State of AI Citations 2026.

Build a competitive prompt set

Cover the questions buyers ask before they know you and once they’re comparing, and tag each prompt with its group:

Prompt groupExamplesWhat it reveals
Category”Best [category] for [type of customer]”, “Top [category] providers in [market]“Who the engines put on the shortlist
Problem”How do I [problem you solve]?”Who gets named when the buyer doesn’t know the category yet
Alternatives”Alternatives to [competitor]”, “Cheaper alternative to [competitor]“Whether you’re named as the switch option
Comparison”[You] vs [competitor]”, “[Competitor A] vs [Competitor B]“Whose framing the engine adopts, and which sources it trusts for the comparison
Branded”Is [you] good for [use case]?”, “[You] pricing”Whether answers about you are accurate, and whether competitors get inserted
Attribute or location”[Category] with [feature]”, “[Category] near [city]“Gaps on specific features or markets

Twenty to fifty prompts covers most single-brand programs. Write them the way people actually type, then lock the list before the first run, or your Day 30 numbers won’t be comparable with Day 0.

Comparison prompts deserve special attention, because third-party sources tend to win them. In data Ahrefs published, 22 questions naming its own brand were tracked (August 28 to October 4, 2026): for factual questions such as “ahrefs pricing”, an Ahrefs page was cited on 97–100% of days, but for comparisons such as “ahrefs vs semrush”, AI Overviews and AI Mode cited an Ahrefs page on 0–5% of days and ChatGPT on 71% (Ahrefs, October 29, 2026). One brand and 22 questions, but the pattern is worth checking in your own category.

Find the competitors that actually appear

Run the whole set once as a discovery pass and list every brand named. Then fix your tracked competitor set: usually the five to eight brands that appear most often, plus any direct rival you must report on. Keep it fixed between cycles and state it next to every share-of-voice number: adding or dropping a competitor changes your share even when the answers don’t change. Add newcomers only at a re-baseline.

Expect surprises, and expect famous brands to start far ahead. In a June 2026 preprint from the AI visibility tracking company Ranqo (Kumar, arXiv:2606.20065), covering 100,000+ responses about 100+ brands on ChatGPT, Claude, Perplexity and Gemini (March–May 2026), household names appeared in 73% of relevant answers on their first run, mid-market and regional brands in 44%, and niche brands in 11%. Compare yourself with brands of similar size, not only the category leader.

Run the prompts and log brands and sources

AI answers change from run to run, so one pass is a sample. In SparkToro and Gumshoe’s January 2026 study (600 volunteers, 12 prompts, 2,961 runs on ChatGPT, Claude and Google’s AI), the tools returned the same list of brands less than once in 100 runs. How often a brand appeared was far steadier than where, so Rand Fishkin concluded that visibility percentage across many runs is a reasonable metric and “ranking position” is not. (Not peer-reviewed; Gumshoe sells AI visibility tracking.)

Run each prompt three to five times per engine at minimum, in fresh, unpersonalized sessions from a constant location; adding distinct prompts helps more than repeating one. The full sampling method is in How to Measure AI Citations. For competitor analysis, log one row per prompt, engine and run with at least:

  • Brands named, in order of appearance, counting each brand once per answer
  • Domains and URLs cited, every one, not only yours
  • Whether your site was linked, and which page
  • How each brand is described (“the budget option”, “best for enterprise”), copied exactly
  • Answer type: shortlist, comparison, single-brand description, refusal, or no AI answer shown

A spreadsheet works for a first pass. At hundreds of rows per cycle an AI visibility tool saves time (Best AEO Tools in 2026); pick one that exports raw answers with their cited sources. First-party data won’t show competitors: Bing Webmaster Tools’ Citation Share (preview since June 16, 2026) gives your share of citations for a grounding query, but Microsoft says it “does not expose competitor domains.”

Score appearance rate, citation rate and share of AI voice

Use the same definitions as How to Measure AI Citations:

  • Appearance rate = answers that name the brand ÷ total answers sampled
  • Citation rate = answers that link to the brand’s site ÷ total answers sampled
  • Share of AI voice = the brand’s mentions ÷ mentions of all tracked brands (yours included)

Here’s an illustrative example: 100 answers on one engine, four tracked brands.

BrandAnswers naming itAppearance rateAnswers linking to its siteCitation rateShare of AI voice
You3030%1212%30 ÷ 150 = 20%
Competitor A6060%1818%60 ÷ 150 = 40%
Competitor B4545%3131%45 ÷ 150 = 30%
Competitor C1515%44%15 ÷ 150 = 10%
Total150 mentions———100%

Appearance rates don’t add up to 100% because one answer can name several brands; shares of voice always do. Report both, by engine and prompt group. Here A is named most but B’s own pages are cited most, suggesting B wins on its own content and A on what other sites say about it, which call for different responses.

No published cross-industry benchmark exists for a “good” share of AI voice, so compare yourself with your own baseline and the tracked set. And mind the margin of error: at a 30% appearance rate, 100 answers give roughly ±9 points at 95% confidence (1.96 × √(0.3 × 0.7 ÷ 100)), and repeated runs of one prompt aren’t fully independent, so the real uncertainty is wider.

Map the sources cited instead of you

This is the part most competitor reports skip, and it produces the work list. Take every answer where a tracked competitor appears and you don’t, list the domains cited, and group them by type:

Source typeExamplesWhat it usually meansTypical response
Competitor’s own pagesTheir comparison page, pricing page, guideTheir page answers the question more directly or completely than yoursImprove or create your equivalent page
Review and listing sitesG2, Capterra, Trustpilot, Google Business Profile, industry directoriesThe engine trusts aggregated customer opinion or directory data for this questionComplete and correct your listings; ask real customers for reviews
Editorial and “best-of” articlesTrade publications, ranked listiclesThird parties have written about the category and named them, not youEarn inclusion with something reviewers can verify
Forums and communitiesReddit threads, Q&A communitiesReal users recommend them in public discussionsTake part genuinely; make sure your product solves the problems people raise
VideoYouTube reviews and tutorialsReviewers and creators cover themPitch reviewers; publish useful videos of your own
ReferenceWikipedia, standards bodies, government sitesThe question is definitional or factualMake sure your facts are accurate where these sources describe your category

Count how often each domain appears across your gap answers; the few that recur most are your priorities. Three findings help set expectations:

  • The Ranqo preprint found about 78% of citations went to corporate websites, and ranked “best-of” listicles were the most-cited content format, at about 21% of citations.
  • Ahrefs’ November 3, 2026 study of 10,000 US brands and 260 million non-branded prompts across AI Overviews, AI Mode and ChatGPT found every type of off-site mention correlated strongly with how often a brand was named, led by Reddit mentions (0.75 Spearman), then web pages (0.70) and news articles (0.67). Ahrefs stresses that correlation isn’t cause and brand size explains part of the link.
  • Ahrefs’ October stability study suggests that if a YouTube video, Reddit thread or major publisher consistently holds a citation you want in AI Overviews, getting featured in that content may be more realistic than replacing it.

Diagnose why a competitor’s page gets cited

For each competitor page cited in your gap answers, open it next to your closest equivalent and compare:

  • Does it answer the question directly? Pages that state the answer early and plainly are easier for an engine to use.
  • Is it specific and checkable? In the original GEO research (Aggarwal et al., KDD 2024), adding quotations, statistics and citations to credible sources raised a source’s visibility in generated answers by 30–40% on the main metric, while keyword stuffing offered “little to no improvement.” It was a lab benchmark, not a live-engine study; see What is Generative Engine Optimization?
  • Do you have a matching page at all? Often the competitor has a “[Competitor] vs [You]”, alternatives or pricing page and you don’t.
  • Is it more current? Make real updates where your facts are out of date. Date changes alone don’t count: Google lists “changing the date of pages to make them seem fresh when the content has not substantially changed” as a warning sign (helpful content guidance). Method: The AEO Content Refresh Playbook.
  • Who else vouches for them? Count the reviews, comparison articles, forum threads and press mentions naming them versus you. This is often the real gap (Review Management for AI Visibility covers reviews).
  • Schema is hygiene, not the explanation. Ahrefs tracked 1,885 pages that added JSON-LD against about 4,000 matched controls (May 2026) and found no citation gain in ChatGPT or AI Mode and a small decline in AI Overviews. Google says “structured data isn’t required for generative AI search” (Google Search Central). Details: Structured Data and AI Citations.

Turn the gaps into a work list

Group what you found by gap type. Each needs a different fix and often a different owner.

Gap typeWhat you seeFix
Missing pageThe competitor has a page for the question (comparison, alternatives, pricing, definition); you don’tPublish one that answers the question in its opening lines, with specific, sourced facts
Weaker pageYou have a page, but the competitor’s is more direct, specific or currentRewrite the answer, add verifiable figures and sources, update outdated facts
Third-party coverageThe cited sources are review sites, listicles or forums that name the competitor and not youComplete listings, request reviews from real customers, pitch editors with something verifiable, take part in community discussions honestly
Inaccurate descriptionYou’re named, but with wrong prices, features or positioningCorrect your own pages first, then the third-party pages the engine cites
Engine-specific gapYou appear on one engine but not anotherCheck which source types that engine cites (listings, your site, forums) and work on those

Prioritize prompt groups closest to revenue (category shortlists, comparisons with your closest rivals), and fixes that close gaps across several prompts at once, such as one comparison page or one directory listing. Where nobody leads clearly, the clearest answer has the opening. Once fixes ship, re-run the same locked prompts on the same schedule and judge the change against the margin of error. How long changes take to show up varies by engine and source type; the evidence is in How Long Does AEO Take?

Monitoring and defending your share of AI voice

Once you have a position worth keeping, the job is watching it without overreacting to noise.

Tier prompts by value. Check revenue prompts (category shortlists, rival comparisons, branded pricing) most often; review informational prompts monthly.

Sample on a schedule, and expect churn. Ahrefs’ October 2026 stability study followed about 37,500 cited websites on AI Overviews, AI Mode, ChatGPT, Gemini and Copilot for 20 days: 50–83% of citations were gone the next day, but about 75% of those came back within the 20 days, and citations that had lasted eight days or more almost always returned. Brand names recurred far more consistently than cited sites. Ahrefs advises counting how many of the last seven days you were cited and looking at a page only if a citation is still missing after two or three weeks.

Alert on drops that hold. A sensible trigger is a fall in appearance rate or share of AI voice larger than your margin of error that holds for two consecutive cycles, or a tracked competitor newly appearing on revenue prompts.

Don’t let one URL carry everything. If most of your citations point to one page, a single competitor improvement can cost you several prompts. Focused pages for distinct questions spread that risk.

When you lose ground, diagnose before you act. Re-run the prompts, see who appears now and which sources are cited, fix the specific gap, then re-measure. Log what you changed and when, so you can tell your fix from an engine update or seasonal demand.

What not to do

  • Don’t copy a competitor’s page. Google’s guide to generative AI features asks for “valuable, non-commodity content.” A paraphrase of the cited page gives an engine no reason to switch.
  • Don’t fake or buy reviews. The FTC’s consumer reviews rule, in effect since October 21, 2024, prohibits fake reviews, buying positive or negative reviews and certain review-suppression tactics (16 CFR Part 465).
  • Don’t bump dates or rotate prompts to make numbers look better.
  • Don’t report an AI “rank” or judge from one engine. Report rates, engine by engine.

Want a competitor baseline without building it yourself? Our free AI Visibility Snapshot runs 10 prompts on ChatGPT, Google AI Overviews and Perplexity and notes where competitors show up instead of you. The Citation Audit tests 30–50 locked prompts across ChatGPT, Google AI Overviews, Perplexity, Gemini, Claude and the OpenAI Responses API, with competitor source mapping, schema and content fixes and a prioritized 60–90 day roadmap in 5 business days. It costs $1,500, fully credited toward an Implementation Sprint or retainer signed within 30 days. Ongoing competitor citation monitoring is part of the Authority retainer ($4,500/month, 100 locked prompts). We don’t guarantee citations or rankings; we measure the same locked prompts the same way each time so you can judge the trend.

Request a free Snapshot or a Citation Audit → · See pricing


FAQ

What is share of AI voice?

Share of AI voice is your share of all brand mentions across the competitors you track. If 100 AI answers contain 150 mentions of your tracked brands and 30 are yours, your share is 20%. It’s different from appearance rate (being named in 30 of 100 answers is a 30% appearance rate), and it changes if you add or drop competitors, so keep the tracked set fixed and state it with the number.

How do I find out which competitors ChatGPT recommends instead of me?

Write 20 to 50 prompts your buyers would ask, including “best [category] for [use case]”, “alternatives to [competitor]” and comparison questions. Run each several times in fresh sessions with search on, and record every brand named and every source cited. Repeat on the other engines that matter to you, because each engine leans on different sources.

Why is my AI competitor different from my Google competitor?

AI engines cite many pages that don’t rank in Google’s top 10 for the query, each engine favors different sources, and AI search leans heavily on third-party coverage. A competitor that’s widely discussed elsewhere can be recommended more often than one with a stronger website.

What is a citation gap?

A citation gap is a prompt where a tracked competitor is named or cited and you aren’t. To close it, find the sources cited in those answers, work out whether the gap is a missing page, a weaker page or missing third-party coverage, and fix that specific cause. Then re-run the same prompts to check.

How often should I check competitor AI visibility?

Weekly checks with several runs per prompt suit the person doing the work, but make decisions on monthly and quarterly trends. Citations churn daily and most that disappear come back, so act on drops that exceed your margin of error and hold for two cycles.

Do I need a paid tool for AI competitor analysis?

No. A spreadsheet and a fixed prompt list are enough for a first baseline. Tools help at hundreds of answers per cycle; our AEO tools comparison lists current prices and engines.


Sources

All sources checked November 11, 2026.

Related reading: How to Measure AI Citations, Best AEO Tools in 2026, State of AI Citations 2026, Review Management for AI Visibility, The ROI of GEO.