Keyword Research in the AI Era: Intent Beats Volume
For years, keyword research meant one thing: find terms with high search volume and tolerable competition, then write pages to match. That playbook is breaking — not because search is dying, but because a growing share of answers now come from AI Overviews and assistants instead of blue links.
Here’s the uncomfortable part: for small businesses, the volume-chasing playbook was already a losing game. Now it’s worse. The way out is the same in both cases — stop optimizing for volume, start optimizing for intent.
Why chasing volume was always a trap — and is now a bigger one
A head term like “social media marketing” has enormous volume and near-zero attainability for a small agency. The first page belongs to global publishers and software companies with years of accumulated authority. That was true before AI.
What’s new: generic informational queries are exactly the ones AI Overviews and chat assistants answer directly. When someone asks “what is social media marketing?”, the answer appears above the results — synthesized, cited, done. Even if you somehow ranked, the click increasingly doesn’t happen.
The independent data points the same way. Pew Research Center tracked the browsing of 900 US adults in March 2025: on Google visits that showed an AI summary, people clicked a regular result 8% of the time, against 15% when there was no summary, and clicked a link inside the summary itself on just 1% of visits (Pew Research Center). In a vendor study built on aggregated Search Console data, SEO tool maker Ahrefs found that by December 2025 the presence of an AI Overview correlated with a 58% lower click-through rate for the top-ranking page (Ahrefs).
So high-volume generic keywords now combine the worst of both worlds: you probably can’t rank, and if you did, the traffic is evaporating. The queries that still send visitors — and customers — are the specific ones.
From keywords to questions
People don’t type fragments into assistants; they ask full questions. “Best way to promote a community event in Vancouver on a small budget” is a query now, not a conversation. Your research should mirror that shift.
The richest question sources aren’t in any SEO tool. They’re in your own business:
- Sales emails and DMs — the questions prospects ask before they buy are the questions others are asking a search engine or an assistant right now.
- Customer conversations — objections, confusions, “how does X work with Y” moments.
- Reviews — yours and your competitors’. Review language is buyer language, verbatim.
Keep a running list of these questions. It will outperform any keyword tool export, because every entry comes with proof that a real buyer asked it.
One honest caveat: question-shaped searches are also the ones most likely to get an AI answer. In Pew’s data, 60% of searches that began with “who”, “what”, “when” or “why” produced an AI summary, compared with 8% of one- or two-word searches. So treat your question list as a map of what buyers need to know, and write answers you’d be happy to see quoted. The clicks come from the questions closest to a decision.
The three intent tiers
Every query sits on a ladder of intent:
Informational — “what is…”, “how does…”
The searcher wants understanding, not a vendor. This tier has the most volume and the least commercial value — and it’s the tier AI answers most completely. Publish here selectively, to demonstrate expertise, not to harvest traffic.
Comparative — “X vs Y”, “best… for…”
The searcher is weighing options. “Agency vs in-house social media”, “best marketing approach for a restaurant opening”. Assistants answering these queries name specific options — and searchers still click through to validate. In a July 2026 BrightLocal survey of US consumers, 75% of people who used AI to find a local business checked Google or Maps before contacting it (eMarketer). This is where small businesses start winning.
Transactional — “pricing”, “near me”, “hire”
The searcher is ready to act. Volume is tiny; value is maximal. A page that cleanly answers “how much does social media management cost in Vancouver” will never trend — it will just quietly produce enquiries.
For local businesses, this is also the tier AI has changed least. In Whitespark’s 2025 study of 540 queries across three US cities, AI Overviews appeared on 15% of local-intent searches, while the local map pack appeared on 93% (Whitespark). “Near me” and neighbourhood searches still run through your Google Business Profile and the basics in our local SEO guide. And once those searchers land, the page has to turn them into enquiries, which is what our conversion audit checks.
The strategy writes itself: concede the informational head terms, own the comparative and transactional long tail. There, being genuinely local and genuinely specific is an advantage no large publisher can copy.
One page, one intent — but a whole question cluster
Long-tail doesn’t mean one page per micro-variation. Search engines and AI systems both understand that “how much does an agency cost”, “agency pricing”, and “what should I budget for social media” are one intent expressed three ways.
Build one well-structured page per query intent: a direct answer up top, ## subheadings for the natural follow-up questions, and an FAQ section with self-contained answers. That single page can rank — and be cited — across the entire cluster. What kills a page is the opposite mistake: mixing intents, so a pricing page drifts into a general explainer and serves no one fully.
This matters more with AI search in the mix. Google says AI Overviews and AI Mode may use a “query fan-out” technique, running several related searches across subtopics to build one answer (Google Search Central). Our reading: a page that already answers the natural follow-up questions gives those searches more to find.
Free research beats expensive tools
You can do all of the above without a subscription:
- Autocomplete — type your service plus “for”, “vs”, “in”, and read what suggests. That’s real query data.
- People Also Ask — expand a few entries and the box refreshes with more; it maps question clusters for you.
- Google Search Console — the queries you almost rank for (positions 8–20) are your cheapest wins. Traffic from AI Overviews and AI Mode is counted in the same Performance report, under the “Web” search type. Since August 31, 2026, every site also gets a Generative AI performance report showing which pages appear in those features, but it reports impressions only, by page, country, device and date, with no query list (Search Console Help). The keyword work still happens in the main report.
- Review language — the exact phrases customers use are the phrases future customers search with.
What about Google Keyword Planner? Google calls it a free tool (Google), but you have to finish setting up a Google Ads account, billing details included, before it will suggest keywords (Google Ads Help). It’s useful for seasonality and the rough size of broad demand. It’s weakest exactly where small businesses win: Google notes that keywords with very low search volumes aren’t discoverable in it, and that is where a lot of local, question-shaped demand lives. If you do run a small ads test, its search terms report shows the searches that actually triggered your ads, which is the sequence we lay out in Google Ads vs. SEO.
Where this meets GEO
Everything above doubles as preparation for AI search. Question-shaped pages with clean, self-contained answers are precisely what assistants retrieve and cite — the mechanics we cover in our GEO guide and in how AI assistants recommend local businesses. Intent-first keyword research isn’t a separate discipline from GEO; it’s the front half of it.
Google says as much. Its guidance for AI features states that “optimizing for generative AI search is optimizing for the search experience, and thus still SEO”, with no special files or markup required (Google Search Central). The same question list feeds both.
Want a second pair of eyes on which questions your business should own? Email us and we’ll take a look. Bilingual keyword and content strategy is part of our local SEO and AI search service.
Frequently Asked Questions
Is keyword research still worth doing now that AI answers so many searches?
Yes — but the unit of research has changed from keywords to questions and intents. AI Overviews absorb generic informational queries, so the opportunity has shifted to specific comparative and transactional questions where searchers still click through, and where a small business can genuinely be the best answer.
What is the difference between informational, comparative, and transactional intent?
Informational queries seek an explanation ("what is event marketing"), comparative queries weigh options ("agency vs freelancer for social media"), and transactional queries signal readiness to act ("social media agency Vancouver pricing"). For small businesses, comparative and transactional long-tail queries convert far better and face far less competition than informational head terms.
Where can I find keyword ideas without paying for a tool?
Four free sources cover most needs: search engine autocomplete, the People Also Ask box, your own Google Search Console query report, and the exact language customers use in reviews, emails, and sales calls. Together they surface the real questions your buyers ask, in the words they actually use.
Can Search Console show me which of my keywords trigger AI Overviews?
Not by keyword. Since August 31, 2026, every site has a Generative AI performance report in Search Console, but it shows impressions in AI Overviews and AI Mode broken down by page, country, device and date, with no query list. Traffic from those features is also counted inside the regular Performance report under the Web search type, so do your query research there and use the new report to see which pages AI features are picking up.
Keep reading
Get in touch