How to Identify Search Intent by Reading the SERP

The fastest way to identify search intent is to search the keyword yourself and read the first ten results as evidence: whatever format Google already ranks is the format it has decided satisfies that query. If eight of ten results are listicles with "best" in the title, your 2,000-word philosophical essay is unlikely to rank, however good it is.

That is the whole method in two sentences. The rest of this article is the detail: what to look for, how to record it, what the odd cases mean, and how to turn a SERP reading into a brief you can actually hand to a writer or an AI tool without getting a mismatched draft back.

Search intent is not something you guess from the words in a keyword. "Best CRM" and "CRM software" look similar and want very different pages. You read intent off the results page, because the results page is a record of what Google has already tested against real user behaviour. You are not predicting. You are reading a report.

Why guessing intent from keyword text fails

Most founders classify keywords by eyeballing them. The keyword contains "how to," so it must be a guide. It contains "vs," so it must be a comparison. That heuristic works often enough to feel reliable, but the cases it gets wrong are frequently the keywords worth the most.

A few examples of how the heuristic breaks:

• "Email marketing" looks informational. In practice the SERP is often dominated by product pages and category pages from ESPs, because most people typing it want a tool, not a definition.
• "How much does a freelance SEO article cost" looks like a simple factual question. The pages that rank tend to be long, with tables of price ranges by writer tier, because the searcher is trying to build a budget, not learn a number.
• "Notion templates" looks like a listicle query. Often it returns a mix of gallery pages, marketplace pages, and a couple of blog roundups, meaning a plain article is fighting a directory.

The four classic buckets (informational, navigational, commercial investigation, transactional) come from search literature and are still useful as labels. Google's own Search Quality Rater Guidelines use a related framing of "know," "do," "website," and "visit-in-person" needs. But labels alone do not tell you what to write. "Commercial investigation" could mean a nine-item listicle, a two-product comparison, or a calculator. You need the format, not just the bucket.

So the method below is format-first. You read the SERP, you name the page type, and then you build the brief around that page type.

Prerequisites: what you need before you start

You do not need paid tools for this. You need:

1. A keyword list. Ideally already grouped. If you have a raw dump from a keyword tool, run it through a grouping pass first; the keyword clustering guide covers turning a list into a publishing plan so you are not reading SERPs for forty variants of the same query.
2. A clean browser session. Personalised results skew what you see. Use an incognito or private window. If your audience is not in your country, add a location parameter or use a VPN, because SERPs differ by market more than people expect.
3. A spreadsheet with six columns. Keyword, dominant format, dominant angle, average depth, SERP features, decision. That is your whole apparatus.
4. Twenty to thirty minutes. A careful SERP read takes three to five minutes per keyword once you know what to look for. Do a batch of five or six in a sitting.
5. Honesty about your site. A domain with eleven posts and no backlinks reads the same SERP differently from a domain with four hundred. Part of intent reading is deciding whether you can serve the intent competitively, and that depends on where you stand. If you are early, the piece on how long a new blog takes to rank sets expectations honestly.

Optional but useful: a free Google Search Console property for your site, so you can check later whether your reading was right.

Step 1: Search the keyword in a clean, correctly located session

Photo: Microbiz Mag · BY

Open a private window. Type the exact keyword, not a variation. Do not add your brand. Do not autocomplete into something similar. The keyword you research must be the keyword you write for, or the rest of the exercise is fiction.

Two settings matter:

Location. Google serves different results by country and sometimes by city. If you sell to US buyers and you search from Lisbon, you may see a Portuguese-flavoured SERP with different competitors. You can force a country by appending &gl=us to the search URL, or by setting your region in Google's search settings. For a US reading, the URL pattern looks like google.com/search?q=your+keyword&gl=us&hl=en.

Personalisation. Even in incognito, Google uses some signals. It is not perfectly neutral. That is fine. You are reading patterns, not measuring rankings to two decimal places. If eight results share a format, that pattern survives personalisation noise.

Scroll the full first page. Not the first three results. The first three often include an outlier that Google is testing or an authority site that ranks on brand strength alone. Positions four through ten are frequently more informative about what the query actually wants, because they got there on relevance rather than domain weight.

Verification for this step: you can state the country and language you searched in, and you have looked at all ten organic results plus whatever features sit above and between them.

Step 2: Name the dominant page format

This is the core move. For each of the top ten organic results, assign one label. Keep the label list short and stable so your notes stay comparable across keywords.

A working taxonomy:

Count the labels. If six or more of ten share a label, you have a dominant format and your decision is easy: match it or skip the keyword.

If the split is 5/5 or 4/3/3, you have a mixed SERP, which is its own signal and gets its own step below.

Two practical notes:

Judge by the page, not the title. Titles lie. Plenty of pages titled "The Complete Guide to X" are actually listicles with a paragraph of intro. Open a few and check the actual structure. You do not need to open all ten; open the top three and any that look ambiguous.

Watch for Reddit and forums. When a significant chunk of page one is Reddit threads, Google is signalling that it believes users want lived opinion rather than a polished vendor explainer. That is a hard SERP for a company blog to win with a corporate-sounding article. It is winnable with a genuinely specific, opinionated piece with real numbers, but it is not winnable with a bland overview.

Verification for this step: you have a count like "7 listicles, 2 comparisons, 1 product page" written down.

Step 3: Read the angle inside the format

https://www.youtube.com/watch?v=zXPZ0L2cOw4

Video: "Keyword Research Pt 1: How to Analyze Searcher Intent - 1.2. SEO Course by Ahrefs" — Ahrefs (YouTube)

Format tells you the shape. Angle tells you the contents. Two listicles for the same keyword can be completely different articles.

Look at the top five titles and note:

Qualifier words. "Free," "for small teams," "2026," "open source," "no code," "for beginners." If three of five titles say "free," the searcher's constraint is price, and a roundup of paid enterprise tools will bounce. If titles say "2026" or a current year, freshness is part of the intent and an undated article looks stale.

Item count. If the ranking listicles carry 10 to 15 items, a five-item post looks thin. If they carry seven, a forty-item post looks like scraped filler. Match the range, do not chase the maximum.

Who the pages are for. A "best project management software" SERP full of enterprise review sites is answering procurement buyers. The same keyword with "for freelancers" in half the titles is answering solo users. You cannot write for both in one piece without diluting it.

What is being compared to what. On "X alternatives" queries, note whether the ranking pages compare on price, on features, or on use case. If every ranking page leads with pricing tables, pricing is the decision axis for that audience. This is exactly what shapes a piece like Scalenut alternatives for small teams: the SERP tells you whether readers want a feature matrix or a plain "which one should I actually buy" answer.

Write the angle as one sentence, in plain language. For example: "Twelve to fifteen free or cheap tools, aimed at solo users, with pricing stated per tool and a short verdict each." That sentence is worth more to a writer than a keyword and a word count.

Verification for this step: you have a one-sentence angle statement that a stranger could act on without seeing the SERP.

Step 4: Measure depth, not just format

Depth is the third dimension. Same format, same angle, wildly different effort required.

Check three things:

Word count, roughly. You do not need a tool. Open the top three results, scroll, and estimate. A page that takes four scroll-flicks is different from one that takes twenty. If you want numbers, most browsers let you select all and paste into a word counter, or you can use a free SEO browser extension. Do not treat the average as a target to beat by 20%. Word count is not a ranking factor; it is a proxy for how thoroughly the topic gets covered at this level of the SERP. If the top pages cover eleven subtopics, you need to cover roughly those eleven, in however many words that takes.

Subheading list. Copy the H2s from the top two or three results into your notes. This is the single highest-value artifact from a SERP read. The union of those H2s is the topic's expected coverage. Anything that appears in all three is effectively mandatory. Anything that appears in none is your opportunity to add something.

Evidence type. Do ranking pages carry screenshots, pricing tables, code blocks, original data, embedded video? A SERP where every result has a comparison table is telling you a table is expected. A SERP where every result has annotated screenshots means text alone will underperform, and that is a real cost consideration for automated content.

Verification for this step: you have a merged H2 outline from the top results and a note on required evidence types.

Step 5: Log the SERP features and what they take from you

The organic ten are not the whole page. Features change how much traffic there is to win, and sometimes change what page type wins.

Note which of these appear:

• AI Overview. If Google generates an answer above the results, simple factual queries lose a lot of clicks. This does not mean skip the keyword; it means favour keywords where the reader needs to see options, prices, or step detail that a summary cannot fully replace. The dynamics are covered in more depth in how AI Overviews change blog traffic.
• Featured snippet. Tells you the answer format Google wants: paragraph, list, or table. If the snippet is a numbered list of steps, put a clean numbered list near the top of your page.
• People Also Ask. Free subtopic research. Expand each question, note the answering page, and add the good ones as H3s. PAA questions also reveal intent drift, which matters for step 6.
• Shopping / Ads density. Heavy ads usually mean transactional intent and commercial competition. A blog post competing with four shopping ads and three product pages is fighting uphill.
• Video carousel. If video sits high, the task is likely visual (software walkthroughs, physical processes). A text article can still rank, but embedding a relevant video is more than decoration here.
• Local pack. The query has geographic intent. Unless you are local, skip.
• Site links / brand dominance. If one brand owns most of page one, the query is navigational and you are not the destination.

Verification for this step: your spreadsheet row lists every feature present, in order of vertical position on the page.

Step 6: Detect mixed intent and decide how to handle it

Sometimes the SERP does not agree with itself. Four guides, three product pages, two forum threads, one comparison. This is not noise. It usually means one of three things.

The query is genuinely ambiguous

"Python" means a language, a snake, and a comedy troupe. Google hedges by serving all of them. You will never satisfy this query fully, and you should not try. Go find a more specific keyword.

The query sits mid-funnel

"Marketing automation" attracts people learning the concept and people ready to buy a tool. Google splits the page. Here you have a real choice: write the informational piece and win the guide slots, or write the commercial piece and fight the vendors. For most small sites, the informational slot is the winnable one, and you convert with in-content links rather than by trying to rank a landing page you cannot support with links.

Google is still testing

New or fast-moving topics show unstable SERPs. If you check the same keyword two weeks apart and the format mix has shifted, Google has not settled. This is an opportunity if you can publish quickly and a risk if you spend three weeks on one piece.

A practical rule for mixed SERPs: write for the majority format, and satisfy the minority with a section. If seven results are how-to guides and three are comparison pages, write the how-to and include one section that names the main tool options. You catch the primary intent and cover the secondary without splitting the page's purpose.

What you should not do is create two pages for the same mixed intent and hope one wins. That is how you end up with two of your own pages splitting signals for one query, which is the problem described in keyword cannibalization on a small blog.

Verification for this step: for any keyword you marked mixed, you have written down which format you are targeting and what you are doing about the secondary intent.

Step 7: Check whether the intent is winnable for your site

Reading intent correctly is useless if you cannot execute against it. Two checks:

Format feasibility. If the winning format is an interactive calculator and you write blog posts, you have a build task, not a writing task. Be honest and either build it or move on. If the winning format requires original data (surveys, benchmark tests), you need the data. Some intents are genuinely closed to you right now.

Competitive weight. Look at who ranks. If positions one through ten are all major review sites and category leaders with hundreds of referring domains, a new blog is very unlikely to break in on a general keyword. Look for the longer, more specific version of the query where the SERP includes smaller sites, forum threads, or pages that are clearly thin. A page one containing two Reddit threads and a 600-word post from a site you have never heard of is a page one with a gap in it.

A rough scoring rule you can apply in ten seconds per keyword:

If you are early and finding that everything looks unwinnable, that is normal and it is a sequencing problem, not a strategy problem. The advice in how to start a SaaS blog when you're the only one applies: start with the queries where you have unusual first-hand knowledge, where the competition is thin because nobody else can write the page.

Verification for this step: each keyword row has a decision: write now, write later, or skip.

Step 8: Convert the reading into a brief

Photo: Generationbass.com · BY

The reading is only worth something if it changes what gets written. Turn each SERP read into a short brief. These fields are enough:

Keyword: how to identify search intent
Format: how-to (7/10 on page one)
Angle: practical, step-based, for people doing it manually
       without enterprise tools
Depth: 2,000-3,000 words, 8-10 H2 sections in top results
Must cover: intent types, SERP features, mixed intent,
       turning reading into a brief
Must include: at least one table, a worked example
Do not: open with dictionary definitions of the four
       intent types; every ranking page already does that

That last line matters more than it looks. Once you know what all ten results do, you also know what is saturated. The differentiator on a well-covered keyword is rarely a new subtopic; it is usually a sharper version of an existing one, or a concrete example where competitors gave a generalisation.

If you use an AI writing tool, this brief is the input that determines whether the draft is usable. Generic prompts produce generic drafts because the model has no idea what page type the SERP wants. Feeding it the format, angle, depth and the merged H2 list is the difference between an article that could rank and an article that reads like every other article. This is the main practical argument in is AI-generated content bad for SEO: Google's stated position is about quality and usefulness, not about how the text was produced, and intent match is a large part of what "useful" means in practice.

Verification for this step: the brief is specific enough that two different writers would produce structurally similar articles from it.

Step 9: Publish, then check your reading against reality

SERP reading is a hypothesis. Search Console tells you if it was right.

Eight to twelve weeks after publishing, open the Performance report, filter to the page, and look at the queries. Three outcomes:

Impressions and clicks growing, queries match your target. Your reading was correct. Note what you did and repeat it.

Impressions but almost no clicks, average position 15-40. You are relevant but not competitive, or your format is close but not right. Check the SERP again. Did the format shift? Are you a guide on a listicle SERP?

Impressions for the wrong queries. Your page matched a different intent than you targeted. This happens when the angle was off. Often the fix is retitling and restructuring the top third of the page, not rewriting the whole thing.

One caution on timing: do not judge at four weeks. New pages on small domains commonly take two to four months to settle into a stable position, sometimes longer in competitive niches. Judging too early leads to rewriting pages that were about to work, which is one of the reasons founders abandon their blog after a handful of posts.

Verification for this step: you have a dated note per page recording predicted intent versus observed queries.

Worked example: three keywords, three different pages

Abstract method, concrete application. Here is how three related keywords resolve into three different page types.

"best cms for saas blog"

Format read: listicles and comparisons dominate. Titles carry three to five named platforms. Some results are vendor comparison pages.

Angle: decision-stage. The reader has narrowed to a handful of options and wants a recommendation with reasons, not a survey of thirty CMSs.

Depth: long, with tables. Pricing appears on almost every ranking page.

Decision: write a comparison of the two or three platforms the audience actually considers, with a table and a clear verdict per use case. Which is roughly the structure of Webflow vs WordPress vs Ghost for SaaS blogs. A generic "what is a CMS" explainer would be format-mismatched and would struggle to rank.

"internal linking for blogs"

Format read: how-to and guide, mixed. Some results are conceptual explainers, some are process posts.

Angle: practitioners looking for a repeatable system, not a definition of what a hyperlink is.

Depth: medium to long, with examples. Screenshots common but not universal.

Decision: how-to with a named system, because the mixed guide/how-to SERP favours whichever page is more actionable. That is the reasoning behind a simple internal linking system.

"seobot alternative"

Format read: comparison pages and vendor pages, heavy brand presence.

Angle: the reader has used or evaluated a specific tool and is looking for a replacement. They want differences, pricing, and a straight answer, fast.

Depth: medium. Bloated comparison posts underperform here because the reader is impatient.

Decision: direct head-to-head with pricing stated and honest limits acknowledged. A SEObot vs Rankody comparison is a legitimate page type for this SERP; a 3,000-word philosophical piece about content automation is not.

Same niche, three keywords, three formats. That is the point. Intent is per-query, not per-topic.

Troubleshooting: common problems and what they mean

"Every result is a huge brand and I have no chance"

Your keyword is too broad. Add a qualifier and re-search. "Project management software" is a brand keyword; "project management software for two person agencies" may have a page one with actual gaps. The narrower query usually has a clearer, more specific intent too, which makes the brief easier to write.

"The SERP is all Reddit and Quora"

Google has decided the query wants human opinion. You can compete, but only with a page that reads like a person who has actually done the thing: specific numbers, named tools, stated tradeoffs, and a willingness to say "this one is not worth it." Polished neutral prose usually loses these SERPs.

"The top result is a page that looks worse than what I'd write"

Check three things before you assume you can beat it: how old it is, how many other pages on that site cover the topic (topical authority), and whether it has links. Sometimes a thin page ranks on domain strength alone and you genuinely cannot displace it with a better page. Sometimes it ranks because it is the only page that matches intent exactly, and yours will beat it. Reading the H2s tells you which.

"I see different results than my colleague"

Location, language, and personalisation. Compare the URL parameters you each used. For a shared reference, agree on one country and language setting and use the URL pattern from step 1 so results are reproducible.

"The format changed between when I researched and when I published"

Happens on volatile topics, especially anything AI-related. Re-check the SERP right before publishing on any keyword you researched more than a month ago. It takes two minutes. If the dominant format shifted, adjust the intro and the structure rather than scrapping the draft.

"My page ranks but gets no clicks"

Two likely causes. First, an AI Overview or featured snippet is absorbing the answer, which is more common on definitional queries than on comparison or how-to queries. Second, your title and description do not match the intent even though the page does. Look at your Search Console CTR against the position-average and rewrite the title to match the angle the ranking pages use.

"I read the intent right but the article still feels generic"

Format match gets you eligible. It does not get you chosen. Once you know the required structure, the differentiator is specificity: a real price, a real screenshot, a real limitation you are willing to name. If your draft could have been written by someone who has never used the product or done the task, the intent match will not save it. This is where a review pass matters, and the pre-publish SEO checklist covers the mechanical part of that pass.

"I have 200 keywords and this method takes too long"

It does, at three to five minutes each. Two mitigations. First, cluster before you read, so you are reading one SERP per cluster rather than per keyword; a cluster of twelve close variants usually shares an intent. Second, batch by likely format: run all the "how to" queries in one sitting, all the "best" queries in another. Pattern recognition speeds up dramatically when you are comparing similar SERPs back to back.

Where automation fits, and where it does not

Reading a SERP is one of the tasks that has held up well against automation, because the judgment call, "is this page actually a how-to or a listicle wearing a how-to title," is still easier for a human skimming than for a scraper counting headings. What automates well is everything downstream: pulling the merged H2 outline, tracking word counts, monitoring whether a SERP's format mix shifts over time, and turning an agreed brief into a full draft.

That split matters when you evaluate tools. A tool that generates 100 articles from 100 keywords without a per-keyword intent decision will produce a lot of format-mismatched pages, which is a common failure mode of bulk generation. The tradeoff is laid out in Byword vs Rankody: volume without an approval step means nobody catches the mismatch until the pages have been live for months. A broader survey of the category sits in best AI blog automation tools.

Rankody's own approach is to do the SERP and topic research as part of planning, then produce drafts that the founder approves before they publish. The approval step is where a format mismatch gets caught, and it takes a couple of minutes per article rather than the hours it costs to fix after publication. That is the honest version: automation handles the volume, a human handles the judgment, and the judgment is mostly this: does this page do what the SERP says the query wants?

A short checklist you can actually use

Photo: Kristin Hardwick · CC0

Print this, or keep it in a note.

1. Search the exact keyword in a private window with the right country parameter.
2. Label all ten organic results by page type. Count the labels.
3. If one format has six or more, that is your format. If not, note it as mixed and pick the majority.
4. Read the top five titles for qualifiers, item counts, and audience.
5. Copy the H2s from the top two or three results. The union is your expected coverage.
6. Note every SERP feature and its vertical position.
7. Judge winnability: are there results from sites your size on page one?
8. Write a six-field brief: keyword, format, angle, depth, must-cover, must-avoid.
9. Publish. Note the date.
10. Check Search Console at week eight to twelve. Compare observed queries to predicted intent.

Ten steps, roughly five minutes per keyword once it is habit. It is the cheapest quality control in content marketing, and it is the reason some blogs publish twenty articles and get traffic while others publish sixty and get nothing.

The pages that fail rarely fail because the writing was bad. They fail because they answered a question nobody was asking on that particular results page.