Research & briefs
How to read a SERP before you write a single word
The results page is a published specification of what Google currently believes the query deserves. Reading it takes ten minutes and settles four decisions.
7 min read
Every query already has an answer sheet: the page of results it returns today. It tells you what page type wins, how deep the coverage goes, which surfaces take the clicks before the blue links start, and whether an independent site can realistically get in.
Most SERP analysis stops at "look at the top three and write something longer". That produces a page that matches the competition on length and on nothing else. A useful pass records specific, comparable observations and ends in four decisions: go or no-go, format, depth, and angle.
What follows is the pass itself, the signals worth recording, how to read the SERP features, and the tells that a query is not worth a page at all.
Set up so you see the real SERP
- Search in a private window, signed out, so personalisation does not reorder the results for you.
- Set the location and language to the market you are targeting; local intent changes the page type entirely.
- Look at the full page before scrolling — what occupies the space above the first organic result is part of the analysis.
- Check mobile as well when the query has a local or transactional flavour; the feature mix differs.
- Screenshot it. SERPs change, and you will want the record when you review performance in three months.
The five signals to record
Record these for the first ten organic results. Ten minutes of note-taking, and the output slots straight into a content brief.
Diversity is the signal most often ignored and the most informative. Ten near-identical guides means the format is settled and your only lever is substance. A mix of a tool page, two guides, a forum thread and a video means Google has not resolved a single intent — which is an opening, because a page that serves the dominant sub-intent well can take a position without out-resourcing anyone.
| Signal | What to note | Decision it drives |
|---|---|---|
| Page type | Guide, listicle, comparison, tool, product, forum, video, documentation | The format your page has to be |
| Result diversity | Are the ten results one type or several? | Whether the query has one intent or a mixed one you can pick from |
| Site type | Vendors, publishers, forums, .gov/.edu, marketplaces | Whether an independent site can compete at all |
| Depth and structure | Heading count, word count range, presence of tables, data, original imagery | Your length budget and the substance required |
| SERP features | AI overview, featured snippet, PAA, video, images, local pack, shopping | Realistic click share, and which format the features reward |
Reading the SERP features
AI overviews and featured snippets
When a generated overview or a snippet sits at the top, the definitional part of the query is already answered before anyone scrolls. A page whose whole value is the definition has little left to offer; a page that answers it in two sentences and then does something the overview cannot — a procedure, a comparison, first-hand data — still has a reason to be clicked.
The way to earn snippet placement is structural: a direct, self-contained answer immediately under a heading that matches the question, in the format the current snippet uses.
People Also Ask
PAA is a free subtopic list. Expand three or four of them, note the questions and the pages Google chose to answer them. Questions that recur across several of your target queries are the ones that deserve their own H2 rather than a passing mention.
Video, image and local blocks
A video block high on the page means a meaningful share of searchers want to be shown rather than told; a text-only page can still rank but will lose clicks. A local pack usually means the query has a geographic intent that a national informational page cannot satisfy.
Ads and shopping units
A dense block of ads and shopping results signals commercial intent regardless of how informational the phrasing looks. It also compresses organic click share, which matters when you are deciding whether the position is worth the work.
Deciding go or no-go
The analysis should be allowed to end the project. These are the patterns that mean the honest answer is no.
The last of them is a judgement call rather than a rule. Big sites rank on generic pages all the time, and a page with genuine first-hand evidence beats a well-resourced page with none. The no-go applies when you have nothing to add — which is worth admitting at the research stage rather than after 2,000 words.
- The whole first page is a brand you are not — a navigational query, not a topic.
- Every result is a product or pricing page and you are planning an article: the intent is transactional and the article will not be shown.
- The results are entirely .gov, .edu or primary sources for a query where authority is the whole answer.
- The results are all forums and community threads: people want peer answers, and a company page will read as the wrong voice.
- Ten near-identical pages from sites far larger than yours, none of which is more than three years old, and you have no distinct angle or data.
Turning observations into an angle
- List what every ranking page includes — This is your table stakes. Missing any of it makes the page feel incomplete, no matter how good the rest is.
- List what none of them includes — Usually a specific number, a worked example, an edge case, a failure mode, or a screenshot of the thing actually being done. This is where an angle comes from.
- Note what they all get wrong — Copied advice propagates. A ranking page repeating a rule that stopped being true is an opening — provided you can cite why.
- Check the dates and the freshness burden — If every ranking page was updated in the last six months, the topic decays and you are committing to maintenance, not just publication.
- Write the angle as one sentence — If you cannot state what your page offers that the current results do not, the analysis is not finished.
Re-reading the SERP after you publish
A SERP pass has a second use that most teams skip. When a page underperforms, compare the current results page to the one you screenshotted. Three outcomes recur, and each has a different fix.
| What changed | What it means | Response |
|---|---|---|
| Page types shifted (guides replaced by tools) | Intent moved; your format is now wrong | Rebuild the format, not the copy |
| The SERP looks the same, you rank 15-30 | Format is right, substance or authority is not | Deepen the page against the table-stakes list, or build links |
| An AI overview or snippet now sits above | Click share fell without a position change | Restructure so the page answers what the overview cannot |
| Your own two pages both appear low | Cannibalisation, not competitiveness | Merge them and redirect the weaker URL |
Questions people ask
How many results should I analyse?
- Ten. The top three show you the winner; results four to ten show you the range of formats Google considers acceptable, which is what tells you whether the pattern is settled.
Does word count from the ranking pages tell me how long my page should be?
- It gives you a budget, not a target. Word count is not a ranking factor; matching the range keeps you from being conspicuously thin, and exceeding it for its own sake produces padding.
Should I still write for a query with an AI overview?
- Yes, if your page offers something a summary cannot — a procedure, first-hand data, a comparison, a tool. No, if the page was only ever going to define the term.
How often should I redo a SERP analysis?
- Before any significant rewrite, and whenever a page loses traffic without losing position. Both symptoms usually mean the results page changed shape rather than your page getting worse.
References
- Google Search Central — Search Essentials — The baseline requirements a page must meet before any SERP-derived decision matters.
- Google Search Central — Featured snippets and how they work — How snippet content is selected, and why structure rather than markup earns it.
- Google Search Central — AI features and your website — Google's guidance on how AI experiences in Search source and link to pages.
- Google Search Central — Creating helpful, reliable, people-first content — The added-value test behind the go/no-go decision and the angle step.
Keep reading
- Naming the intent you just observed — The four categories, the SERP tells for each, and what mismatched intent looks like.
- Recording the pass so it is usable — The brief fields these observations fill: intent verdict, format, subtopics and length budget.
- Writing for the snippet you saw — Self-contained passages that can be lifted as an answer without losing meaning.
- When the SERP moved and your page did not — Deciding between update, merge and prune once a results page has changed shape.