No AEO-native keyword tool exists yet. Ahrefs tells you what people search on Google. It cannot tell you what ChatGPT is answering (or whether your content is being selected as a source). Here is how to build an AEO research process from what the evidence actually supports.
The fundamental problem with using traditional tools for AEO
Traditional keyword research tools — Ahrefs, SEMrush, Moz — measure how often a phrase is entered into Google Search. That measurement is genuinely useful for understanding whether a topic has audience demand, but it answers a different question than AEO requires.
AEO requires knowing whether an AI system generates a clear answer for that query, and whether your content gets selected as a source for it. Google search volume is a proxy for the first half of that question and tells you nothing about the second half.
A peer-reviewed arXiv study found that GPT-4o's source pool overlaps with Google's indexed content by only 4%, and Perplexity overlaps at 15.2% — a gap large enough that high-volume Google keywords predict AI citation eligibility for only a fraction of AI search activity. The rest requires a different research process.
What AEO keyword research is actually trying to find
The output of AEO keyword research is a map of question-intent clusters where AI systems are actively generating answers and your content can become a cited source — not a ranked list of search volumes. Three signals define a good AEO target.
First, the query is answered directly by AI systems when you test it: ask ChatGPT, Perplexity, and Google AI Overviews the question, and if you get a synthesised answer with citations, the query is an active AEO target.
If it declines to answer or returns only links, it is not an AEO target yet.
Second, the intent is informational and specific: question-format queries with clear informational intent — "what is", "how does", "why does", "which is better" — map directly to the retrieval format AI systems use.
BrightEdge research found informational queries appear in AI Overviews at 3 to 4 times the rate of navigational or transactional queries.
Third, the topic has sufficient depth for you to produce a definitive, citable answer — AI systems favour content that reduces their own uncertainty.
A page that answers the question comprehensively with original data, cited sources, or first-party expertise creates a stronger retrieval signal than a page that summarises the same information available everywhere else.
How do you conduct AEO keyword research without a purpose-built tool?
Three data sources combine to answer this without a dedicated tool: Ahrefs or SEMrush for demand signals, direct testing inside ChatGPT and Perplexity for whether AI systems are actually answering the query, and People Also Ask or autocomplete data for how users phrase the question when they type it into a search box.
Each source alone covers only part of the underlying question. Search volume without citation testing tells you a topic is popular, not that an AI system produces an answer for it.
Citation testing without demand data risks optimising for a query almost nobody asks — combining all three is what separates a validated AEO target from a guess.
| Data source | What it tells you | What it cannot tell you |
|---|---|---|
| Ahrefs / SEMrush | Topic demand, competition, search volume trends | Whether AI systems are answering the query or citing sources |
| Direct LLM testing | Which queries are being answered, what sources are cited, how content is summarised | Scale: you can only test queries one at a time |
| People Also Ask / autocomplete | Conversational query variants, question-based intent patterns | AI citation likelihood or source selection criteria |
| SERP AI Overview presence | Which queries Google AI Overviews is active on | ChatGPT, Perplexity, or Gemini standalone behaviour |
Source: BetterAISearch methodology synthesis, May 2026
Step 1: build a topic seed list
Start with the topics your business has genuine expertise in, not just the topics with the highest search volume.
AEO performance is driven by E-E-A-T signals — experience, expertise, authoritativeness, and trustworthiness — and producing content in areas where you have first-hand expertise creates a citation advantage that content in adjacent, volume-driven areas cannot match.
Run your seed topics through Ahrefs or SEMrush filtered to informational intent, looking for queries with meaningful volume, manageable competition, and clear informational intent — this narrows your candidate set to topics where audience demand exists.
Step 2: validate against actual AI systems
Validating against actual AI systems means testing each candidate topic directly in ChatGPT (with browsing enabled), Perplexity, and Google Search with AI Overviews active.
Record three things: whether the AI generates a synthesised answer, whether it cites specific sources, and whether any cited sources are competitors, adjacent sites, or publications you could realistically equal or surpass in authority.
Queries that return active AI answers with citable sources are confirmed AEO targets, while ones that return only links, or where it declines to synthesise, are not yet AEO targets — even at high Google search volume.
AirOps analysis of 815,484 pages found that content appearing in AI citations was structured with7 to 20 subheadings and ran 500 to 2,000 words for per-query citation rate optimisation.
The AirOps finding confirms that content format matters as much as topic, and the format should match the query type: a definitional query needs a clear, short, extractable answer, a comparative query needs a structured comparison, and a how-to query needs sequential steps.
Step 3: map conversational variants
Mapping conversational variants starts from how AI systems actually receive queries: in natural language, not keyword-optimised strings. Users ask ChatGPT "what's the best way to optimise my content for AI search?" not "AI search optimisation tactics".
Your AEO keyword research should capture the conversational variants of each target topic.
People Also Ask and autocomplete in Google are the most accessible sources of conversational query data — Ahrefs' Questions filter within keyword explorer exports PAA data at scale, and AlsoAsked.com maps PAA trees visually.
Both are legitimate AEO research inputs because conversational phrasing predicts how queries will be entered into AI systems.
Is there a purpose-built AEO keyword research tool?
As of mid-2026, no purpose-built AEO keyword research tool exists, though several products are developing in this space. What exists today mostly tracks which sources AI systems cite for a monitored keyword set, functioning as a citation monitor rather than a keyword discovery tool.
The closest existing categories are AI answer monitoring platforms that track whether your domain appears in AI responses to a set of queries, and share-of-voice tools adapted for AI search. These are useful for measuring AEO performance, not for initial keyword discovery.
AEO keyword research in 2026 is still a largely manual process. The combination of traditional keyword tools for demand validation and direct LLM testing for AEO qualification is the most dependable process available right now.
Purpose-built tooling will emerge; the methodology outlined here is what works now.
What to prioritise with your AEO keyword list
Once you have a validated AEO keyword list (topics with demand, active AI answering, and citable source pools), prioritise by three factors.
First, topical authority alignment. Producing content where you have genuine expertise produces better E-E-A-T signals and is more likely to be cited for the long term. A well-researched answer from an identifiable expert outperforms a comprehensive answer from an anonymous domain.
Second, citation gap analysis. If the current cited sources for a query are weak (thin content, unattributed claims, outdated data), there is a genuine citation opportunity. When the cited sources are Wikipedia, academic papers, or official platform documentation, the bar is higher.
Third, content freshness potential. Perplexity weights content published or updated in the last 30 days most heavily. Topics where you can sustain regular updates (because new data emerges, platform behaviour changes, or research accumulates) have compounding AEO value over time.
The most reliable AEO process combines keyword tools with direct LLM testing
Keyword tools alone identify topic demand; they don't identify AI citation eligibility, which is why pairing them with direct LLM testing and conversational query mapping produces the most dependable AEO research process available today.
The research is clear that question-based informational queries with expert-attributed, structured content drive the highest AI citation rates.
The methodology to find those queries exists today: it is just manual. Purpose-built AEO keyword tools will catch up, but this combination of demand data and direct testing is what produces results in the meantime.