An AEO competitor analysis tells you why AI engines recommend other brands instead of yours, and exactly what to fix. Unlike traditional SEO, you are not chasing keyword rankings — you are reverse-engineering which sources, claims, and content patterns models pull into their answers.
Step 1: Define the Competitive Set
Start by separating who you think you compete with from who AI actually surfaces.
- Direct competitors — the brands you sell against in deals.
- AEO competitors — whoever co-appears with you in AI answers, including review sites, listicles, and aggregators that may outrank every vendor.
Run 20-40 buyer-intent prompts (“best [category] for [use case]”, “alternatives to [you]”, “is [competitor] worth it”) across ChatGPT, Perplexity, Gemini, and Google AI Overviews. Record every brand and domain that appears. The brands that show up most often — not your sales-deck rivals — are your real AEO competitors.
Step 2: Measure Share of Voice
Quantify the gap before you try to close it. Reading the answers yourself, log for each prompt whether each competitor is mentioned, their position in the answer, and the sentiment. Aggregate into a share-of-voice score per engine. See share of voice explained for the scoring method and competitor benchmarking for tiering.
Build a simple matrix:
| Prompt | You | Competitor A | Competitor B |
|---|---|---|---|
| best tool for X | not cited | #1 | #3 |
| alternatives to A | #2 | n/a | not cited |
Patterns jump out fast: maybe a competitor owns every “best for enterprise” prompt while you only appear in “cheapest” prompts.
A per-competitor position column like the one above is something you fill in by eye, and it is the one column automated tracking does not keep for you. What it keeps is your position among all the options an answer named, plus which of your configured competitors were named alongside you — co-appearance, not their placement. Keep the manual column while the running order is what you are arguing about; once you move to the tracked set in Step 6, read your own position against a co-appearance count instead.
Step 3: Reverse-Engineer Their Citations
This is where the real intel lives. For prompts where a competitor wins and you lose, click into the cited sources (Perplexity and AI Overviews show them directly; for ChatGPT, ask it to list sources).
Catalog for each winning answer:
- Which domains are cited — their own site, G2, Reddit, an industry blog, a news outlet?
- What page type — a comparison page, a docs page, a third-party review?
- What specific claim the model lifted (“supports SSO on all plans”, “rated 4.8 on G2”).
You will usually find competitors win because of third-party validation and direct, extractable claims — not prettier marketing copy. That matches the finding in the original GEO: Generative Engine Optimization paper, which tested content changes against generative engines and found that adding quotations, statistics and cited sources raised a page’s visibility, while stylistic changes such as keyword stuffing did not. Use win/loss analysis for AI visibility to formalize this comparison.
Step 4: Run a Gap Analysis
Now translate findings into a backlog. Cross-reference where competitors are cited against where you are absent. A full method is in the content gap analysis guide.
Look for three gap types:
- Source gaps — they are on G2, Reddit, and Capterra; you are not. Citation and review presence is often the biggest lever.
- Content gaps — they have a “[category] for [vertical]” page answering a prompt you have nothing for.
- Claim gaps — they make specific, structured claims AI can extract; your pages bury the same facts in prose.
Step 5: Identify Their Weaknesses
Competitor analysis is not only about catching up. Find prompts where everyone is weakly cited or where AI relies on outdated information about a rival. Those are openings to publish definitive, fresh content and claim the answer outright. Understanding how AI recommends products helps you spot which decision criteria are still up for grabs.
Step 6: Prioritize and Track
Score each opportunity by prompt volume (how often buyers ask it), competitive intensity (how entrenched the leader is), and effort. Attack high-volume, low-intensity prompts first.
Then re-run the same prompt set on a fixed cadence. AEO is a moving target — model updates and new competitor content shift answers constantly, so a one-off analysis is stale within a quarter.
This is the part worth automating, and automating it means narrowing the set: promote the prompts you actually intend to defend onto the project’s tracked set — up to 25 active prompts per project on a paid plan, 10 on Free — and let competitor benchmarking re-run that fixed set every refresh. The 20–40 you swept by hand is research; the tracked set is the instrument, and it has to be small enough to hold constant. Your overall share of voice is the figure that gets plotted over time; the per-engine split is read inside whichever scan you open, so use it to spot where a change landed, not as a trend line per engine.
Quick checklist
- [ ] 20-40 buyer-intent prompts logged across 4+ engines
- [ ] Share-of-voice matrix built per engine
- [ ] Cited sources catalogued for every competitor win
- [ ] Source, content, and claim gaps documented
- [ ] Opportunities scored and added to a backlog
- [ ] Monthly re-run scheduled
Frequently Asked Questions
How is AEO competitor analysis different from SEO competitor analysis?
SEO analysis focuses on keyword rankings and backlinks for your own pages. AEO analysis focuses on which sources AI engines cite when recommending a category — often third-party sites like Reddit and G2 rather than vendor pages. The output is a list of citations and claims to win, not keywords to rank for.
Which AI engines should I include in the analysis?
At minimum cover ChatGPT, Perplexity, Google AI Overviews, and Gemini, since they drive the most buyer research. Add Claude and Copilot if your audience uses them. Answers vary significantly between engines, so analyzing only one gives a misleading picture.
How often should I repeat a competitor analysis?
Run a full analysis quarterly and a lightweight share-of-voice check monthly. Model updates and new competitor content can reshuffle answers within weeks, so continuous monitoring beats one-off audits. Two of the automatic alerts do part of this watching for you — a competitor spike when a tracked rival gains 10 or more points of share of voice since your previous scan, and a visibility score change of 5 points or more in either direction. Neither is scoped to a prompt, and there is no alert for losing a citation you used to have, so the quarterly read is still where a narrower shift gets caught.
What if a review site outranks every vendor in AI answers?
Treat the review site as a primary AEO channel, not a competitor to beat. Invest in your profile, gather reviews, and ensure your listing carries the structured claims AI extracts. Winning the aggregator often moves your visibility more than improving your own site.