Reading search intent from the SERP itself
Stop guessing intent from the keyword alone
I used to sort keywords into buckets by looking at the word itself. “Best” meant commercial. “How to” meant informational. “Buy” meant transactional. That system works about as often as it fails, and the failures are expensive because you find out after you’ve published 2,000 words that don’t match what’s actually ranking.
The keyword tells you what someone typed. It does not tell you what they wanted. The SERP tells you what they wanted, because Google has already run the experiment for you across millions of clicks and dwell times. Every result on that page is a vote for a specific interpretation of the query. Your job is to read the votes before you write anything.
This is the part of search intent analysis most people skip. They’ll run a keyword through a tool, get a volume number and a difficulty score, and start drafting. I run the keyword myself, look at the actual page, and only then decide what to build.
What the SERP features are already telling you
Before you even look at the ten blue links, look at what’s sitting above and around them.
A shopping carousel means Google has decided a meaningful slice of searchers want to compare products right now, not read about them. If you see product listings with prices and star ratings at the top, writing a 3,000 word buying guide as your primary asset is fighting the page, not working with it. You might still want that guide, but it’s not what wins the click.
A local pack (the map with three business listings) means the query has a “near me” component even if the person didn’t type “near me.” I’ve watched sites throw content at queries like this for months with nothing to show for it, because no amount of on page optimization fixes a query Google has classified as local intent. That’s not a ranking problem, it’s a category mismatch.
Featured snippets tell you Google found a page that answers the question in a compact, extractable way. Look at what’s snippeted. Is it a definition, a numbered list, a table, a single sentence? That’s the format Google’s system rewarded for that exact phrasing. It’s a strong hint about the structure that satisfies the query, not proof that only that structure can rank.
People Also Ask boxes are underused. Each question in that box is a real sub intent tied to the main query. If you’re building a page and PAA is stacked with pricing questions, comparison questions, and “is it worth it” questions, that’s a commercial evaluation intent hiding inside what looked like an informational keyword.
Video results mixed into an organic result set usually mean the query has a demonstration or tutorial component. This shows up constantly on process based queries, anything involving a physical or on screen step by step. If three of the top ten spots are YouTube videos, a text article is competing for a smaller remaining slice of attention than the SERP layout suggests.
Read the actual pages that are ranking
Once you’ve clocked the features, open the top five to eight organic results and look at what kind of page each one actually is. Not the content, the page type.
Is it a single product review? A comparison table of multiple products? A tool or calculator? A definitional glossary entry? A long form guide? A forum thread or Reddit post? A news article?
I keep a simple habit here: I list the page type next to each of the top ten, nothing fancier than a spreadsheet column. When seven or eight of them are the same type, that’s your answer. Google has already run the intent classification and the SERP is the output. Building a different page type into that lineup is a bet against the pattern, and it’s a bet I’ve lost more times than I’ve won.
Mixed pages types in the top ten mean mixed intent, and that’s useful information too. If you see two comparison pages, three single product reviews, a listicle, a definitional page, and a forum thread, the query doesn’t have one clean intent. It has several audiences searching the same phrase for different reasons. In that case the winning move is usually to build a page broad enough to serve the dominant type while covering the secondary intent as a section within it, not to pick one and ignore the rest.
Titles and meta descriptions are intent fingerprints
Read the titles in the SERP as a set, not one at a time. If eight of ten titles include a year, that’s a freshness signal Google is rewarding, and it usually means the underlying content needs regular updates to stay relevant, not a one time publish. If titles lean heavily on “vs” or “comparison,” the searcher is choosing between named options, not learning a concept from scratch. If titles are phrased as direct questions, the query is likely informational and probably answerable in a snippet, which changes how much content you actually need on the page versus how much is just padding for word count.
I stopped chasing arbitrary word counts a while back for exactly this reason. If the ranking pages are 600 words because the query only needs 600 words to answer, writing 2,500 doesn’t make you more helpful, it makes you slower to load and harder to scan.
SERP volatility as a signal, not noise
If you track rank positions over time, or even check a query on different days, you’ll sometimes see the SERP shuffle heavily, different domains rotating through the top spots week to week. That volatility itself is information. It usually means Google hasn’t converged on a single best answer for that query yet, which often happens with newer topics or ones where user satisfaction signals are still thin. Volatile SERPs are riskier to target because the goalposts are still moving, but they’re also more winnable because the incumbents haven’t proven they’re unbeatable. Stable SERPs, where the same five sites have held the same five spots for a year, are telling you the opposite: satisfaction signals are settled and dislodging them takes a genuinely better page, not a slightly better one.
A workflow that actually holds up
Here’s what I do before greenlighting a page, on my own sites, every time:
Search the exact keyword in an incognito window from a location that matches your target audience. Screenshot or note every SERP feature present. List the page type of each of the top eight to ten organic results. Read the top three PAA questions and note whether they shift the intent you assumed from the keyword alone. Check whether titles cluster around a format, a year, a comparison structure, or a question. Only after all of that do I decide what the page should be.
This takes fifteen minutes per keyword if you’re doing it properly, longer if the SERP is mixed and you need to decide how to split coverage. It is not glamorous. It will not make a video thumbnail. But it’s the difference between publishing content that matches what Google has already validated searchers want, and publishing content that matches what you assumed they wanted.
What this doesn’t fix
Reading the SERP tells you what to build, not whether you can rank it. A page that perfectly matches intent still needs the site authority, the links, and the technical health to actually surface. I’ve built intent perfect pages on new sites that sat on page four for months, because intent match was never the bottleneck for that domain. Treat SERP reading as the first filter that stops you from wasting effort on the wrong content type, not as a shortcut past the rest of the work.
It also won’t give you a timeline. Search intent shifts as Google’s systems keep testing what satisfies a query, so a SERP you read today is a snapshot, not a permanent map. Recheck it before a major content refresh, especially on commercial queries where product results and shopping features move fastest.
If you want more of this kind of breakdown, walked through on live SERPs instead of described in the abstract, that’s what we build out on the channel and the rest of the site. Head back to the homepage for the full library.