Why this signal matters for Google
A query is not a topic, it is a request. Someone types "running shoes". They might want to buy a pair today. Or compare two models before choosing. Or simply find out after how many miles you replace them. Three people, three different pages, one query. Determining search intent means picking which of those three people you serve, before you write a line. Skip that choice and you publish a page that covers the topic without solving anyone's problem. It is the most common failure I see in audits: the writing is solid, the main content quality is there, and the page never moves.
Google grades this separately. Human evaluators rate pages using a public manual, the Quality Rater Guidelines (QRG for short). They give two ratings that do not stand in for each other. The first judges the quality of the page. The second, called Needs Met, judges one thing only: does this result solve the problem of the person who typed that query, on their phone, right now. Sections 13.0 and 14.0 describe that whole job. They teach raters to read the request behind the words first: does the person want to know something, do something, go somewhere, or reach one specific site. They then hand them the scale, running from Fully Meets, the complete and immediate answer, down to Fails to Meet, the result that is useless for that query. A page can be flawless and land at the bottom: it answers a different question.
Missing the intent costs more on sensitive topics. When a page touches health, money or safety, Google files the topic under YMYL, for Your Money or Your Life, and section 3.4.1 expects real trustworthiness there. Take someone searching "bank transfer stuck 5 days". They want to know what to do now and who to call. Serving them a long piece on the history of payment systems misses the request at the exact moment it matters. The right page puts the essentials at the top, then goes deep for whoever wants the detail.
AI engines make the gap more visible. When ChatGPT, Perplexity or Google's AI summaries build an answer, they do not lift a whole page: they take the passage that settles the question. A page that buries its answer under 1,200 words of preamble has no passage to offer. A photographer who writes "a wedding is 8 hours on site and 400 edited photos delivered within 3 weeks" gives a quotable sentence. The competitor who writes "every project is unique" gives none. It is the same work as for generative search: answer first, elaborate second.
The NeuroEEAT scoring grid (0-10)
Rate, page by page, the gap between the real request behind the target query and what your page actually hands the visitor.
| Score | Description | Concrete indicators |
|---|---|---|
| 0-2 | Off the request | The page covers the topic and solves nothing. The visitor goes back to the results and opens a competitor. No intent analysis happened before writing. |
| 3-4 | Answer buried | The right information exists, but it sits behind the company story and three paragraphs of preamble. On a phone, the reader gives up before reaching it. |
| 5-6 | Partial answer | The main question is handled and the page stops there. The obvious follow-up questions are nowhere, so the visitor opens a second result to finish the job. |
| 7-8 | Immediate and complete | The request is identified, the answer lands in the first screen, follow-up questions are covered further down. The page format matches what the query calls for. |
| 9-10 | Intent verified and maintained | The intent assumption has been tested against the live results, each secondary intent has its own page, and the page is revisited whenever the request or the shape of the results changes. |
How to audit it on your site
Five checks to run on your ten most visited pages, one query at a time.
- 01
Type the query like a stranger
Open a private window and type the query the page targets. Look at the top ten results and write down what they are: guides, product pages, comparisons, videos, brand pages. That list tells you what Google has already worked out about the request. If your page is the lone comparison among ten product pages, that is not originality, it is a misread.
- 02
Read what Google shows around the results
Featured snippet, related questions, local map, image or review blocks: each one reveals the shape of the request. Related questions that all start with "how much" mean the person wants a price range, not a definition. A map block means they intend to go somewhere.
- 03
Time how long it takes to reach the answer
Open your page on a phone and count the swipes before the answer appears. Past a screen and a half, you are making someone wait who already has nine other results in hand. Raters judge the result as it presents itself on mobile, not the desktop version.
- 04
List the follow-up questions
Once the main answer is given, ask what the person looks for next. After "how much does a move cost" comes "what drives the price up" and then "how far ahead should I book". If those follow-ups are not on the page, the visitor goes to find them elsewhere, and stays there.
- 05
Spot pages chasing two requests at once
A page that tries to sell and explain in the same screen serves both badly. Find the pages on your site that mix a guide with a sales pitch, then split them into two pages joined by a link. Each one picks up a clear request, and you stop competing with yourself on the same query.
How to activate or improve it
Three levels to move from a page that covers a topic to a page that settles the request.
Identify the request before writing
- Write in one sentence what the person wants to get, starting with a verb: understand, compare, buy, call, download.
- Test that sentence against the live results, not against your own hunch: if the top ten do not do what you pictured, you are the one who is wrong.
- Put the direct answer in the first two paragraphs, ahead of any context or brand story.
- Give the page a title that echoes the words of the query, so the reader recognises their own request at a glance.
Match the format to the request
- Adopt the format that dominates the results: a comparison table if the person compares, a step-by-step if they want to do something, a short factual page if they want one number.
- Add a follow-up section at the foot of the page, built from the related questions Google shows and from what customers ask you on the phone.
- Split mixed intents into separate pages and link them, so each page serves a single request.
- Check, in your rank tracking tool, which queries the page already shows up for: they describe the real request better than any assumption.
- Make sure the first mobile screen contains the answer, not just a hero image and a menu bar.
Keep intent alive over time
- Record, for every important page, the request it serves and the date you last checked it against the results.
- Re-run the query every six months: when the shape of the results changes, the request has changed, and the page has to follow.
- Watch the pages that collect impressions without clicks: that is the clearest sign of a title promising one thing and a page delivering another.
- Break your pages into blocks that each answer one precise question, with a subheading that states the question: that is what makes a passage quotable by AI engines.
- Write down who, on your team, signs off on the intent of a page before it goes live.
5 common mistakes to avoid
Sector specifics
Medical information site
Someone typing a symptom usually wants one thing: should I see someone, and how soon. Section 3.4.1 expects real trustworthiness on these pages. Put the warning signs and the referral advice at the top, then go into depth. An article that opens with anatomy answers a question nobody asked.
Online store
On a query containing "best" or "comparison", the person is not ready to buy: they are shortlisting. A product page misses the request, and an honest comparison page takes the spot. Keep the product page for queries that name the exact model, where the request is clear and immediate.
Software company
Problem queries and solution queries do not serve the same person. "Our invoices go out late" comes from a manager trying to understand. "Automated invoicing software" comes from someone already comparing tools. Two pages, two formats, two calls to action.
Trade or neighbourhood business
Many local queries want an immediate action: call, drive over, check an opening time. The request is the go-somewhere type, one of the cases described in sections 13.0 and 14.0. A tappable phone number, an address and a price range at the top of the page beat every paragraph of company presentation.
Signals to activate in parallel
Needs Met is built with the signals that keep the answer solid, current and quotable.
FAQ
How do you determine search intent when a query is ambiguous?
Look at the top ten results: when Google mixes several formats for one query, it has spotted several requests and is hedging. Pick the one you can serve best, make it your main page, and handle the second on a linked page rather than inside the same text.
What is the difference between search intent and Needs Met?
Search intent is the person's request. Needs Met is the rating human evaluators give a result against that request, described in sections 13.0 and 14.0 of the QRG. One describes what the reader wants, the other measures whether your page delivers it.
Can one page serve two intents at once?
Rarely well. When you try to explain and sell in the same screen, you disappoint both visitors. What works is splitting into two linked pages, each with its own request, format and call to action.
Should you copy the format of the pages ranking first?
Copy the format, not the content. If the top ten are step-by-step guides, the request calls for a step-by-step, and showing up with a personal essay sinks you. Inside that format, your experience and your examples are what set the page apart.
How often should you recheck the intent of a page?
Twice a year on your important pages, and immediately if traffic drops while the ranking holds. Requests evolve: a query that called for a definition two years ago may call for a price comparison today. The shape of the results is your best indicator of that shift.
Anonymized case study
Before · a page that covers the topic
After · a page that settles the request
Illustrative figures meant to show the typical progression of the signal. They do not come from a real client case and must be replaced with your own measurements before any external communication.