Sorting by search volume is a weak way to pick keywords. A more honest sizing model accounts for winnability, intent match, and how much demand AI-generated answers are already intercepting.
The default version of keyword research is volume-first: sort a list of candidate terms by search volume, pick the biggest numbers, write content for them. As a way of allocating a limited content budget, this is a weak model. Volume is one input into whether a keyword is worth targeting. Treated as the only input, it reliably points teams at keywords that look attractive on a spreadsheet and perform badly in practice.
Why Volume Alone Misleads
Three separate problems hide inside a single volume number.
Winnability
A keyword with strong volume and a page-one dominated by domains with a decade of accumulated authority is not the same opportunity as a keyword with identical volume and a weaker competitive set. Volume tells you the size of the prize. It says nothing about your odds of winning it, which depend on the existing competition for that specific query — not your domain's overall strength, but the strength of what's actually ranking there right now.
Intent Mismatch
Volume aggregates every searcher who typed a given phrase, regardless of what they actually wanted. A term can carry meaningful volume while being dominated by an intent your page doesn't serve — informational searchers landing on a term you intended as commercial, or vice versa. Ranking for a keyword whose dominant intent doesn't match your page produces traffic that doesn't convert, which is a worse outcome than not ranking at all, because it consumes content-production budget for a metric (a ranking) that doesn't translate into the business outcome you actually wanted.
Demand Interception by AI-Generated Answers
This is the variable that didn't exist in keyword research five years ago and matters increasingly now: a meaningful share of query volume for informational and definitional terms is being answered directly inside an AI Overview or a conversational AI response, before the searcher ever reaches a results page they'd click through from. A keyword can carry strong historical search volume and still be a mediocre target today if the dominant behavior for that query has shifted toward "read the generated answer, don't click anything." Recognizing which of your candidate keywords fall into that pattern — typically well-established definitional and how-to queries — versus which still reliably send click-through traffic is now part of sizing the opportunity honestly, not a side consideration.
This is also, from the other direction, exactly why GEO work matters independently of traditional click-through volume: if a query is being answered inside a generated response rather than a clicked link, the opportunity isn't zero — it shifts from "rank and get a click" to "be the source the generated answer cites." That's a different kind of visibility, measured a different way; see our guide to tracking AI citations for how to measure it.
A More Honest Way to Size an Opportunity
Rather than sorting by volume alone, weight a candidate keyword against three questions: how large is the addressable volume once you discount for the share likely intercepted by generated answers, how winnable is it given who's actually ranking there today rather than some abstract difficulty score, and does the dominant intent for this specific phrase actually match what you're planning to publish. A keyword that scores moderately on raw volume but strongly on winnability and intent match is very often the better investment than the highest-volume term on the list.
In practice this means looking past the single aggregate volume number and into the composition underneath it — the actual pages currently ranking (are they resourced the way a small team can realistically compete with, or are they backed by domains with entrenched authority), and, where the tooling supports it, some signal of how much of that query's demand is landing on AI-generated answers versus traditional results.
Research Ends Where Tracking Begins
Sizing an opportunity is a one-time (or periodically-refreshed) analytical decision: is this keyword worth the investment. Once you've made that call and published against it, the question changes to an ongoing measurement one: is it actually working. Those are different tasks with different cadences, and conflating them is a common failure mode — teams that treat research as a continuous activity end up perpetually re-evaluating opportunities instead of measuring outcomes. We cover the measurement side, including why a single manual rank check isn't a reliable way to answer that question, in how to track keyword rankings without checking manually every day.
SearchVitals' Keyword Research tool is built around the sizing questions above rather than a bare volume sort — opportunity data alongside the detail view for a given term, so a candidate keyword can be evaluated on more than one number before it becomes a content commitment. Once you've picked targets, the actual on-page execution — where the keyword needs to appear, how the page should be structured around it — is covered in our On-Page SEO guide.