Title Scans Are a Bad Proxy for Buying Intent — Here's What to Read Instead

A buying intent signal is any observable behaviour on a search results page — click share, result composition, feature mix, dwell — that reveals whether the searcher wants to purchase, compare, or just learn. Genuine buying intent signals are not, and have never been, the presence of the word "buy" or "best" in a page title.

That distinction sounds pedantic. It costs money.

The Cheap Trick Everyone Uses

Open a keyword tool. Export the top ten titles. Count how many say "best," "review," "vs," "pricing," "buy." Assign a score. Call it intent. Move on.

Fast. Wrong. Fast and wrong.

The signal a searcher actually sends is compressed into the composition of the SERP itself — the ratio of shopping units to informational blocks, the number of AI Overview citations, whether Reddit and YouTube outrank the vendor pages, whether ads clear the fold, whether the "People also ask" block asks setup questions or comparison questions. The titles are downstream of all that. They rank because Google has already decided what the intent is. Reading them back to infer intent is reading the verdict and calling it the trial.

Andy Crestodina's three-bucket taxonomy — Know, Do, Go — is useful precisely because it forces the analyst to look past the wording of the queries and pages and toward what the searcher wants to happen next. Ahrefs and Semrush both classify intent using SERP feature analysis, not editorial title language. Google Search Central separates content type, content format, and content angle — none of which map cleanly to the words in a ranked title.

If the taxonomy people built their taxonomy from SERP features, the analyst using titles is working two abstractions downstream.

What Buying Intent Signals Actually Look Like

A commercial SERP has a fingerprint. Shopping carousel above the fold. Two or three ad slots that clear meaningful clicks. Product review pages from independent sites clustered in positions 3-7. A "People also ask" block full of comparison questions ("X vs Y," "is X worth it," "how much does X cost"). Thin representation from Wikipedia, Reddit, YouTube. Little to no AI Overview, or an AI Overview that hedges toward retailer citations.

An informational SERP has a different fingerprint. Wikipedia in the top three. YouTube tutorial. Reddit thread. A definition-heavy answer box. An AI Overview citing three or more sources — Pew Research Center's Google AI summary analysis found 88% of AI summaries cited three or more sources, and only 1% cited a single source. Ads that misfire and get ignored.

Semrush's 2025 review of AI Overviews puts numbers on the drift: in January 2025, 91.3% of queries triggering an AI Overview were informational; by October that share had dropped to 57.1%, while navigational triggers climbed from 0.74% to 10.33%. The mix moves. Intent is not a label you paste on a keyword and forget. It is a live composition that shifts as Google's model of the query shifts.

Titles do not carry any of this. Titles are the last thing to update.

Where Title Scans Fail Loudest

Consider a query like "CRM software." A title scan sees "Best CRM software 2025," "Top CRM tools compared," "CRM pricing guide" and shouts commercial. But look at the SERP. If the top three organic slots are dominated by editorial roundups from publishers whose entire model is affiliate revenue, the searcher is being routed into a comparison funnel — that is late-stage research, not a ready-to-buy signal. The vendor pages sit lower. The ads take the actual transactional clicks.

Per First Page Sage's 2026 report, position 1 organic CTR sat at 39.8% in 2025, up from 39.6% the year prior; position 2 moved from 18.4% to 18.7%. Meanwhile paid top-three CTR slid from 1.7% to 1.5%. Ad Position 1 on clean SERPs holds a 2.1% CTR, per Ghose's Stern School analysis; positions 2 and 3 sit at 1.4% and 1.3%. When ads underperform their historical baseline on a query that looks transactional by title, the intent is softer than the titles suggest. Buyers are still shopping, not buying.

And when an AI Overview enters the picture, the click economy collapses further. Ahrefs' study of 300,000 keywords found that in December 2025, an AI Overview correlated with a 58% lower CTR for the top-ranking page against forecast. Position one CTR for AI Overview keywords dropped from 0.073 in December 2023 to 0.016 in December 2025. Pew Research Center's browsing-panel study found users who saw an AI summary clicked a search result in 8% of visits, versus 15% without one — and clicked a link inside the summary itself only 1% of the time.

Title-based intent scoring has no field for any of that.

How SERP Inspection Actually Runs

The work is not glamorous. It is a lot of tabs.

Query capture: Every target keyword gets run in an incognito window against the localised Google endpoint that matches the buyer market — a US searcher and a Singapore searcher see different SERPs for the same string. The full first page gets screenshotted, not just the ten blue links. Feature composition matters more than rank.

Feature audit: Each SERP is coded for the presence and prominence of shopping units, AI Overviews, "People also ask," video carousels, image packs, local packs, sitelinks, and ad slots above the fold. A commercial SERP without a shopping unit is a research SERP wearing commercial vocabulary. A informational-looking query with three ads and a shopping carousel is a buyer query dressed in a question.

Result-source classification: Each of the top ten organic results is tagged by source type — vendor page, independent review, forum, video, publisher roundup, government, Wikipedia, documentation. The mix is the intent. Heavy Reddit and YouTube presence signals unresolved research; heavy vendor and comparison-page presence signals late-stage evaluation.

Click-economy read: The analyst checks whether the SERP even distributes clicks. An AI Overview citing Wikipedia, YouTube, and Reddit — the three most-cited sources in Pew's analysis, collectively 15% of AI summary sources — plus a "People also ask" block that answers the query in-line means the page is a dead end for traffic regardless of ranking. The intent may be commercial; the opportunity is not.

Cross-market check: Because the same keyword resolves differently across the US, UK, Ireland, Australia, Singapore, and UAE English SERPs, any keyword that will be targeted in more than one market gets inspected in each. Semrush's data showed AI Overviews peaking at nearly 25% of keywords in July 2025 before sliding to 15.69% in November — the ground is not stable, and a market read from six months ago is a market read from a different SERP.

Why People Keep Scanning Titles Anyway

Because reading is expensive.

Jakob Nielsen's essay on why web users scan — first published in 1997, updated through 2010 — puts the cost plainly: reading from screens runs about 25 percent slower than reading from paper. A 2005 update noted that Microsoft's ClearType lifted screen reading speeds by 5%. Not enough to close the gap. Enough to explain why every analyst, given a list of keywords and a deadline, will scan titles rather than open twenty SERPs.

The problem is that the reader-scanning research was about users foraging for information, not analysts assessing markets. When Nielsen's users scanned, the cost of a bad read was five wasted seconds. When a marketing lead scans titles and mislabels a keyword cluster as transactional, the cost is a quarter of misaligned content and a page that ranks for the wrong intent — or ranks and doesn't convert.

There is a legal analogy worth pausing on. In Clapper v. Amnesty International (2013), the U.S. Supreme Court held that fear of surveillance was too speculative to count as injury in fact. Spokeo v. Robins later tried to clarify what harm actually confers standing. Both cases, cited in Solove and Citron's work on data breach harms, wrestle with the same problem the intent analyst faces: the visible artefact (a title, a policy, a breach notice) is not the harm or the signal. The harm is downstream, probabilistic, and only legible if you look at behaviour. Title scans are the standing-doctrine version of intent analysis. They mistake the artefact for the substance.

A Comparison Worth Lifting

Below: what a title scan tells an analyst versus what SERP inspection tells them, drawn from the figures in the research above.

Signal	Title scan reads	SERP inspection reads	Source anchor
Position 1 organic CTR (2025)	Not measurable	39.8%, up from 39.6%	First Page Sage 2026 report
Paid top-3 CTR shift (2025)	Not measurable	Fell from 1.7% to 1.5%	First Page Sage
AI Overview presence effect	Invisible	58% lower CTR to top page (Dec 2025)	Ahrefs, 300,000-keyword study
Click on any result when AI summary present	Invisible	8% of visits vs 15% without	Pew Research Center
Informational share of AI Overview triggers	Assumed static	Fell from 91.3% (Jan) to 57.1% (Oct) 2025	Semrush Report 2025
Ad Position 1 CTR (clean SERP)	Not measurable	2.1%	Ghose, Stern School

Every row in that table is a signal that title text does not carry. Every row is something a searcher's actual behaviour on the SERP does carry.

What This Means for Keyword Research

Buying intent signals live in the composition of the results page and the way clicks distribute across it, not in the words on the ranked titles. That is the whole argument.

The practical read: any keyword research process that ends at a spreadsheet of keywords with intent tags derived from title text is producing a document that looks like analysis and functions like a guess. The keyword list has to be re-graded against SERP features, against source-type mix, against the presence and citation behaviour of AI Overviews, and against the market the searcher is actually in. That work does not scale by scanning faster. It scales by inspecting more SERPs, more often, and treating the SERP as the primary document — the keyword string as just its title.

Matt Kenyon's "Content Well" advice on protecting a business from AI Overviews — repurpose long-form video into shorts, transcripts, and social posts — makes sense only after the intent read is right. Diversifying distribution for a keyword you have misclassified as commercial is diversifying the wrong bet across more channels.

Read the SERP. The titles are the last thing that matters.

Sources

• Do people click on links in Google AI summaries? — Pew Research Center
• Why Web Users Scan Instead of Reading — Jakob Nielsen, NN/G
• Semrush Report: AI Overviews' Impact on Search in 2025 — Jana Garanko
• Google Click-Through Rates (CTRs) by Ranking Position — First Page Sage
• Analyzing the Relationship Between Organic and Sponsored Search — Ghose, NYU Stern
• 3 Types of Search Intent — Andy Crestodina, LinkedIn
• Risk and Anxiety: A Theory of Data Breach Harms — Solove & Citron
• SERP Intent Analysis: How to Decode What Google Really Wants — CONTADU

FAQ

Why are title scans a bad proxy for buying intent?

Because titles are downstream of Google's intent decision, not evidence of it. Reading them back to infer intent is reading the verdict and calling it the trial. The actual signal lives in SERP composition — shopping units, ad slot performance, PAA question types, and whether Reddit and YouTube outrank vendor pages.

What should I read instead of page titles to gauge intent?

Read the SERP itself as the primary document. Code each result page for feature composition, source-type mix across the top ten, ad clearance above the fold, AI Overview citation behaviour, and PAA question shape. If PAA asks setup questions the intent is soft; if it asks "X vs Y," the buyer is late-stage.

How does an AI Overview change the intent read on a SERP?

It collapses the click economy and rewrites what "ranking" is worth. When Wikipedia, YouTube, and Reddit dominate the citations, the page is a dead end for traffic regardless of position. The intent may still be commercial, but the opportunity isn't — a distinction title scans cannot register.

Do I need to inspect SERPs separately per market?

Yes. A US searcher and a Singapore searcher see different SERPs for the same string, so any keyword targeted across the US, UK, Ireland, Australia, Singapore, or UAE gets inspected in each localised endpoint. A market read from six months ago is a market read from a different SERP.

What does a commercial SERP fingerprint actually look like?

Shopping carousel above the fold, two or three ad slots clearing meaningful clicks, independent review pages clustered in positions 3-7, PAA full of comparison questions, and thin Wikipedia, Reddit, and YouTube presence. Strip the shopping unit out and you have a research SERP wearing commercial vocabulary.

Why do analysts keep scanning titles if it's wrong?

Because reading is expensive and deadlines are cheap. Nielsen's research pegged screen reading roughly 25 percent slower than paper, and ClearType only clawed back 5%. Given a keyword list and an afternoon, every analyst will scan titles rather than open twenty SERPs — and mislabel a quarter of content as a result.