Knowing how to research keywords is only half the job. The harder question — and the one most businesses get wrong — is deciding which keywords to actually pursue. In 2026, that decision is more complex because you are optimising for two fundamentally different systems: traditional search engines that rank pages, and AI search engines that cite sources inside generated answers.
Finding keyword opportunities gives you a long list. This guide gives you the framework to shorten that list to the keywords that deliver results across both channels.
Key Takeaways
- Keywords must now be evaluated on dual criteria — traditional search ranking potential plus AI citation potential — because a keyword with 10,000 monthly searches delivers zero AI visibility if AI engines answer it from training data without citing anyone.
- The five selection criteria are: search volume and trend direction, AI citation potential, intent alignment across both channels, content gap and differentiation opportunity, and competitive feasibility in both Google and AI.
- Commercial investigation keywords ("best X for Y") sit at the highest-value intersection — they rank in Google, trigger AI Overviews, and consistently generate AI citations.
- Score each candidate keyword across the five criteria on a 1-3 scale; keywords scoring 12-15 are highest-priority targets, 8-11 are solid secondary targets.
- Keyword selection is not a one-time exercise — revisit priorities quarterly as AI citation patterns evolve and search volume trends shift.
Why Keyword Selection Criteria Have Changed
Traditional keyword selection used three variables: search volume, keyword difficulty, and relevance. That framework assumed one channel — Google organic search — and optimised for rank position.
AI search engines introduced a parallel channel with different rules. ChatGPT, Perplexity, Gemini, Claude, Google AI Overviews, Grok, DeepSeek, Microsoft Copilot, and Meta AI do not rank pages — they synthesise answers and cite the sources they trust most. A keyword with 10,000 monthly searches and low difficulty might earn you a page-one ranking, but if AI engines answer that query from training data without citing anyone, the keyword delivers zero AI visibility.
The reverse is also true. A long-tail question with modest search volume can drive significant value if AI engines consistently cite external sources when answering it. The structural difference between AI and traditional search means that keywords must now be evaluated on dual criteria — not just how well you can rank, but how likely you are to be cited.
Five Criteria for Choosing Keywords That Work Across Both Channels
1. Search Volume and Trend Direction
Search volume still matters — it tells you how many people are looking for something. But in 2026, the direction of the trend matters more than the absolute number.
A keyword with declining volume may signal that the query is migrating to AI search. Gartner predicts a 25% drop in conventional search volume by 2026, meaning some keywords that look strong historically are actually losing their audience to AI platforms.
How to evaluate:
- Check 12-month volume trends, not just current snapshots
- Compare Google Search Console click-through rates — declining CTR on stable volume often means AI Overviews are absorbing clicks
- Prioritise keywords with stable or growing volume that also have AI citation potential (see criterion 2)
2. AI Citation Potential
This is the new dimension most teams skip entirely. AI citation potential measures the likelihood that an AI engine will cite an external source — rather than answering from training data alone — when responding to a query.
High citation potential signals:
- The query requires specific, current data or statistics that AI engines cannot generate from training data alone
- The query compares products, services, or approaches — comparison queries demand cited sources for credibility
- The query asks about events, trends, or developments from the current year — AI training data has a cutoff, forcing citation of fresh sources
- AI Overviews appear for the query — research shows AI Overviews now appear on over 13% of searches, with queries of 8+ words having a 57.3% chance of triggering one
Low citation potential signals:
- Simple factual queries with stable, universally known answers
- Queries where the answer fits in one sentence
- Navigational queries where people search for a specific website
Run your candidate keywords through multiple AI platforms. If three out of five cite external sources when answering, the keyword has high citation potential. Citation analysis across platforms reveals exactly where external sources get mentioned and where AI engines rely on internal knowledge.
3. Intent Alignment Across Both Channels
The same keyword can signal different intent in traditional search versus AI search. A user typing a keyword into Google may want a list of options to browse. The same user asking an AI engine may want a definitive recommendation.
Evaluate intent for both channels:
- Informational intent works well across both channels — educational content ranks in Google and gets cited by AI engines when it is comprehensive and well-structured
- Commercial investigation ("best X for Y") is where the biggest dual-channel opportunity exists — these queries rank in Google, trigger AI Overviews, and consistently generate AI citations
- Transactional intent ("buy X," "pricing for X") performs well in traditional search but rarely earns AI citations — AI engines typically do not cite sources for direct purchase queries
- Navigational intent is irrelevant for AI optimisation — skip these entirely
The highest-value keywords for dual optimisation sit at the intersection of informational and commercial investigation intent. These are the queries where people are researching before making a decision — and where both Google and AI engines surface third-party content.
4. Content Gap and Differentiation Opportunity
A keyword is only worth pursuing if you can create content that adds something the existing results do not. This was true for SEO alone — it is doubly true when AI engines are choosing which sources to cite.
How to assess the gap:
- Search the keyword in Google and review the top 10 results. Are they generic, outdated, or missing a specific angle?
- Run the query through ChatGPT, Perplexity, and Gemini. Study the answers. Are they complete, or do the AI engines hedge, qualify, or give vague responses?
- Check if existing cited sources include original data. If no one is providing first-party data, that is your opening — proprietary data creates defensible content that AI engines must cite because the information exists nowhere else
The strongest keyword opportunities are those where both the SERP and the AI answers have clear gaps — where you can rank in Google by being more comprehensive AND earn AI citations by providing data or analysis no one else offers.
5. Competitive Feasibility Across Both Channels
The final criterion filters out keywords where the competition is too strong to be practical. This means evaluating competitors in two separate arenas.
Traditional SEO feasibility:
- Keyword difficulty score — can your domain authority realistically compete?
- Current ranking position — are you already on page one, or starting from scratch?
- Content quality of current top results — can you genuinely create something better?
AI citation feasibility:
- Who is currently being cited by AI engines for this query? If it is exclusively major publications or government sources, earning citations will be difficult regardless of content quality
- How many sources does the AI cite? Queries where AI engines cite 3–5 sources are more accessible than those where they cite only one authoritative source
- Competitive analysis across AI search platforms reveals which competitors already dominate AI citations for your target keywords — and where they are weak
A keyword that is moderately competitive in traditional SEO but wide open for AI citations is often the best investment. You gain search traffic from rankings while building AI visibility that competitors have not yet contested.
Putting It Together: A Practical Scoring Approach
For each candidate keyword, score it across the five criteria on a simple 1–3 scale:
| Criterion | 1 (Low) | 2 (Medium) | 3 (High) |
|---|---|---|---|
| Volume + trend | Declining volume, low absolute | Stable volume | Growing volume or high stable |
| AI citation potential | AI answers without citing | Mixed citation across platforms | Consistent citations across 3+ platforms |
| Intent alignment | Navigational or pure transactional | Informational only | Commercial investigation or dual informational |
| Content gap | Saturated — no clear angle | Some gaps in existing content | Clear gaps in both SERP and AI answers |
| Competitive feasibility | Dominated by high-authority sites in both channels | Competitive in one channel, open in the other | Accessible in both channels |
Keywords scoring 12–15 are your highest-priority targets. Keywords scoring 8–11 are solid secondary targets. Below 8, either deprioritise or find a different angle.
The scoring is intentionally simple. The goal is not precision — it is consistent decision-making that prevents you from chasing high-volume keywords with no AI future or niche AI topics with no search demand.
Three Mistakes to Avoid
Choosing keywords for one channel only. A keyword that ranks well but never gets cited by AI engines has a shrinking ceiling. A keyword that AI engines love but no one searches for has no scale. The best keywords work across both — evaluate accordingly.
Ignoring AI citation patterns. Many teams still select keywords without ever checking how AI engines answer those queries. Run every candidate keyword through at least three AI platforms before committing to it. Prompt research and keyword research are now complementary processes — one reveals what people type into Google, the other reveals what they ask AI.
Treating keyword selection as a one-time exercise. AI search is evolving rapidly. The citation patterns of six months ago may not reflect today's reality. Revisit your keyword priorities quarterly — re-check AI citation patterns, review volume trends, and monitor your AI visibility to see if your chosen keywords are delivering results across both channels.
Start With the Keywords That Matter Most
The best keyword strategy in 2026 is not the longest — it is the most deliberate. A focused list of 20 keywords that score well across all five criteria will outperform a list of 200 selected on volume alone.
Frequently Asked Questions
What is AI citation potential and how do I measure it?
AI citation potential measures the likelihood that an AI engine will cite an external source rather than answering from training data alone. To measure it, run your candidate keywords through ChatGPT, Perplexity, and Gemini. If three out of five platforms cite external sources when answering, the keyword has high citation potential. Queries requiring current data, product comparisons, or year-specific trends have the highest citation potential.
Should I prioritise keywords with high search volume or high AI citation potential?
Neither in isolation. A keyword with high search volume but no AI citation potential has a shrinking ceiling as users shift to AI platforms. A keyword with high AI citation potential but no search volume has no scale. The best keywords score well on both dimensions — use the five-criteria framework to find the intersection.
How often should I reassess my keyword priorities for AI search?
Reassess quarterly at minimum. AI citation patterns shift as models update their retrieval sources and retrain on new data. The keywords that earned citations six months ago may not produce the same results today. Re-check AI citation patterns, review volume trends, and monitor whether your chosen keywords are delivering results across both Google and AI platforms.
Check your current AI visibility to see which keywords already connect your brand to AI search results — and where the biggest dual-channel opportunities are waiting.
