Search Intent Analysis: How to Read a SERP, and How to Audit Intent on Pages You Already Have

Search intent analysis is usually taught from a diagram: four intent types, a decision tree, a framework with a memorable acronym. However, this refresh works from two primary sources instead. Google's quality rater guidelines describe the query categories Google's own evaluators use. Meanwhile, Google's documentation explains what the Search Console query table does and does not contain. Both change the advice.

The audit half of the guide is demonstrated on this page's own data, and the demonstration is unflattering. Over three months this article recorded 129 impressions and no clicks, at an average position of 34. That number is an artefact. Once the junk queries are removed, the position on queries a person would type is closer to 70.

The previous version of this page claimed seven "proven" steps, invented a framework called I-MATCH, and left six sections as empty headings. It also stated that intent-matched pages "generate the behavioural signals Google interprets as quality: time on page, low bounce rate, return visits". Google has never documented those as ranking signals. Consequently, that sentence is gone along with the rest.

What search intent is, in Google's own words

Search intent is what a person is trying to accomplish when they search. The useful question is not "what did they type" but "what would count as a good outcome for them".

Most SEO writing sorts queries into four buckets: informational, navigational, commercial investigation, transactional. That taxonomy is an industry convention, and it is fine as shorthand. However, it is not Google's. Google's Search Quality Rater Guidelines, updated 11 September 2025, tell raters to think of queries as having one or more of these intents [1]:

Two details in the guidelines matter more than the labels themselves.

First, Know Simple is a narrow box. Google's definition: a query qualifies if "most people would agree on a correct answer, and it would fit in 1-2 sentences or a short list of items" [1]. The guidelines then list what is not Know Simple. That list includes broad or in-depth informational queries, ambiguous queries, queries with no definitive right answer, and queries where "users want to browse or explore a topic" [1]. Most of what bloggers target sits in that second list. Accordingly, writing a 40-word definition and expecting a featured snippet is optimistic.

Second, Google tells its own raters not to force the categories. On Do and Know queries the guidelines say: "Don't worry about strictly differentiating between these two categories. They're given to illustrate some big-picture patterns in user intents, and many queries do not fit neatly into one and only one of these categories" [1]. There is also a whole section on queries with multiple user intents [1]. Therefore, if the people paid to classify queries for Google are told the boundaries blur, a content brief demanding a single confident label is overfitting.

Why intent decides format, without the folklore

The mechanism is simpler than the usual explanation. Google's systems aim to satisfy the query, and the current SERP is the visible output of that aim. A page whose format contradicts the SERP is arguing with the evidence. Similarly, Google's helpful content guidance puts the point from the content side: write for people, and make sure a reader leaves "feeling they've learned enough about a topic to help achieve their goal" [4].

What the mechanism is not is a set of engagement metrics. Time on page, bounce rate and return visits are not documented Google ranking signals, and Google does not receive them for most sites. Intent mismatch does hurt, but through a route you can observe. The wrong format does not get clicked, or gets clicked and abandoned. Ultimately it never accumulates the links and repeat demand that competitive queries need.

The search intent analysis process: five steps, evidence first

Step 1: Read the query and write down a guess

Before opening a search box, decide what you think the searcher wants and which category it falls into. Write it down. Consequently the SERP can contradict it, which is the most useful thing that can happen in this process.

Step 2: Search it clean

Private window, target country, no personalisation. If your readers are in more than one market, run it twice. The intent Google infers is not always the same across locales, which the rater guidelines address directly when they discuss user location [1].

Step 3: Read the SERP as data, not as competition

Work down the first ten results and record, for each:

• Page type: guide, listicle, comparison, product, tool, forum, documentation, video.
• Format signals: does it lead with a definition, a table, a step list, a price?
• Who publishes it: vendors, publishers, forums, official docs. A SERP full of forum threads is telling you people want opinions, not an explainer.
• SERP features: a paragraph featured snippet means a short definition wins the top slot; a list snippet means steps; a table snippet means a comparison. Shopping and local blocks mean the query leans Do or Visit-in-person.
• People Also Ask: the sub-questions Google thinks accompany this one. These are your H2 candidates.

Whatever dominates is the answer. Not what you would prefer to write, and not what one outlier in position 9 did.

Step 4: Classify with a confidence level

Now settle the category, and record how sure you are. High confidence means the SERP is uniform. Low confidence means it is mixed, which is a finding rather than a failure. Mixed SERPs usually want a page that leads with the dominant format and contains the secondary one. Think of a guide with a comparison table in it, or a comparison page with a short explainer at the top.

Step 5: Turn it into a brief

The brief should carry seven things: the confirmed category and confidence; the format the SERP validated; an H2 structure drawn from PAA and from gaps in the ranking pages; a length range taken from what ranks rather than from a target; the outcome a reader should reach; a CTA that matches that outcome; and the internal links that place the page in its cluster. Meanwhile, the competitor content analysis process covers how to extract the ranking pages' structure systematically instead of by eye.

Search intent analysis on pages you already have

This is where most of the value sits, because an existing page has real query data attached and a new page does not. However, it is also where the data will mislead you if you take it at face value.

Open Search Console, filter the Performance report to a single page, and read its queries. Then apply two filters of your own before concluding anything.

Trap 1: anonymized queries

The query table never sums to the page total, and the gap is not a bug. Google's documentation explains that "some queries (called anonymized queries) are not included in Search Console data to protect the privacy of the user making the query". It defines them as queries "that aren't issued by more than a few dozen users over a two-to-three month period". Crucially: "While the actual anonymized queries are always omitted from the tables, they are included in chart totals, unless you filter by query" [2].

For this article, the page total was 129 impressions while the query rows added to 106. Those missing 23 impressions are the rare, long-tail queries. On a small site they are often the most intent-revealing ones you will never see. Expect the gap; do not try to reconcile it.

Trap 2: queries no human typed

The second filter is cruder and more important. Here is the actual query list for this page over three months, grouped by kind:

Thirty-seven of the forty-five queries are numeric-prefixed variants of three strings: segmenting by search intent (19 variants), how to track search intent (16) and search intention (2). No person types 27733: segmenting by search intent. These are automated queries. Moreover, they carry two thirds of the page's listed impressions.

They also wreck the headline metric. Google records position only where a link earns an impression [3]. Those automated queries are so specific that almost anything ranks for them, so they average position 18. Meanwhile, the queries that matter average position 69.7. The blended figure the report shows, 34.4, describes neither.

The audit conclusion for this page is therefore the opposite of what the unfiltered data suggests. It is not a page sitting on page three with a click-through problem. It is a page at roughly position 70 for search intent analysis, which is a visibility problem, and no intent rewrite alone fixes that. Similarly, the content audit guide applies the same filtering discipline across a whole site.

What an intent mismatch actually looks like

Once the query list is clean, the mismatch test is simple. Read the surviving queries and ask whether the page's format is what those searchers wanted.

• Queries are right, format is wrong. Rewrite the page in the format the SERP rewards. Cheapest win in the audit.
• Queries are wrong. The page ranks for things it was not written for. Either re-target it to the demand it actually attracts, or sharpen it and accept it will lose those impressions.
• Queries are right, format is right, position is poor. Not an intent problem. This is the case above, and the work is authority and links, not restructuring.

Mixed intent in search intent analysis, and what to do about it

Plenty of valuable queries return mixed SERPs. Indeed, Google's guidelines expect this and tell raters to "use your judgment when determining if a particular intent is reasonable for queries with multiple intents" [1].

The practical rule: count the formats in the top ten. If seven are guides and three are comparisons, write a guide that contains a comparison, placed where a reader would start asking that question. Otherwise, splitting the page evenly between two formats produces a page that satisfies neither.

Search intent and AI answers

The connection is narrower than most guides claim. AI systems compose answers from pages they can read and quote. Therefore a page that matches intent and states its answer plainly is easier to cite than one that buries it. That much is mechanical.

What cannot be claimed honestly is a reliable method for getting cited. Anobee's own fixed-prompt testing across three runs in September 2026 produced 7, 4 and then 11 cited URLs from Perplexity, and nothing from Google AI Mode on any run. Match intent and write quotable sentences, because both are good practice. Nevertheless, measure citations rather than assuming them, as the guide to checking AI citations sets out.

Common mistakes in search intent analysis

• Classifying from the keyword alone. The wording is a hypothesis. The SERP is the evidence.
• Forcing one label. Google tells its own raters that many queries do not fit neatly into one category [1].
• Auditing intent on unfiltered Search Console queries. Two thirds of this page's listed impressions came from strings no person typed.
• Treating average position as a fact about your target query. It is an average across every query that earned an impression [3].
• Reconciling the query table with the page total. It will never balance; anonymized queries are counted in totals only [2].
• Copying the ranking pages' length instead of their format. Length is a symptom of the format, not the cause of the ranking.
• Citing engagement metrics as ranking signals. They are not documented as such, and the previous version of this page made exactly that claim.

Search intent analysis checklist

Before writing

• Write the intent guess down before searching
• Search in a private window, in the target market
• Record page type and format for the top ten results
• Note the featured snippet shape and the PAA questions
• Note shopping, local, video and forum blocks
• Settle the category and record a confidence level
• Pick the format the SERP validated, not the one you prefer
• Build the brief: structure, length range, outcome, CTA, internal links

Auditing an existing page

• Filter Performance to the page, 3 to 16 months
• Delete numeric-prefixed and scraper-footprint queries
• Expect the row total to fall short of the page total [2]
• Recompute average position across the surviving queries only
• Compare those queries with the page's format
• Decide: rewrite the format, re-target the page, or fix visibility

Bottom line on search intent analysis

Search intent analysis is an evidence exercise, not a taxonomy exercise. Guess, then check the SERP and let it overrule you. Use Google's categories if you want labels, and take the guidelines' own advice that the boundaries blur. When auditing pages you already own, clean the query data before drawing conclusions. Anonymized queries are missing by design, junk queries can dominate the list, and the average position may describe traffic you never wanted. Ultimately, filter first and the diagnosis usually changes.