SERP Analysis

SERP analysis is the practice of reading the search results page for a query before producing content for it. Tools that automate this process are often marketed as SERP analytics platforms, though the underlying practice is the same. The SERP is Google’s working answer to the question “what satisfies this intent?” Reading it carefully reveals what content type, what depth, and what angle the algorithm has settled on, which is the brief any new page has to meet or beat.

Why does SERP analysis matter?

Most ranking failures come from intent mismatch: producing the wrong type of content for the query, regardless of how well the content is otherwise optimised. The SERP shows the right type because Google has already chosen it.

If every result on page one is a comparison article, an attempt to rank a product page on the same query will fail. If every result is a how-to guide, an opinion piece will fail. If every result is a 3,000-word reference, a 600-word summary will fail. The SERP is the brief.

What should you look for in a SERP analysis?

A useful SERP analysis covers:

Content format. Articles, listicles, comparison pages, product pages, category pages, video, tools? Whatever dominates is what Google has decided satisfies the intent.

Content angle. “Best X for Y” pages have a different angle to “How to choose X” pages even though both are listicles. Read the titles and intros for the predominant angle.

Content depth. Word count is a crude proxy. Look at the depth of coverage: how many sub-topics, how much detail per sub-topic, what supporting evidence (data, examples, citations).

SERP features. AI Overviews, featured snippets, People Also Ask, image packs, video carousels, knowledge panels, ad density. Each affects how much organic real estate is available and which angles work.

Domain types. Brands vs publishers vs aggregators vs forums. A SERP dominated by Reddit threads signals user-generated content is being favoured; a SERP dominated by trade publications signals editorial expertise is being rewarded.

Domain authority of ranking pages. What’s the lowest-authority page in the top 10? If it’s a DR 30 niche site, the query is more accessible than tools’ KD scores might suggest.

Recency of ranking pages. When were the top results published or updated? A SERP full of 2018 content suggests freshness is unimportant for the query (or that no one has produced better content recently); a SERP full of 2025 content suggests recency is required to compete.

Title tag patterns. What words appear repeatedly in titles? Those are what Google treats as relevant for the query.

How do you read SERP features?

Each SERP feature signals something about the query and shapes the available opportunity:

AI Overview present. Google has classified the query as informational, with a clear synthesisable answer. Click-through rate to organic is suppressed; citation in the AI Overview becomes a major optimisation target.

AI Mode. Google’s conversational surface sits alongside the results page rather than inside it, and is now available in over 200 countries and territories.¹ It runs its own retrieval, so ranking in the organic top ten does not guarantee you are surfaced there. Treat it as a separate surface to check, not a feature of the page you are already reading.

Featured snippet (no AI Overview). Position zero capture is available. Pages structured around the specific question being answered often win the featured snippet even from below the top organic position.

People Also Ask (PAA). Google is signalling related questions for the query. PAA expansion can lift CTR for pages matching those related questions.

Image pack. Image search is contributing to the query. Pages with strong image SEO can capture that traffic.

Video carousel. Video content is competitive. If you don’t produce video, the carousel reduces available organic real estate.

Knowledge panel. The query is entity-driven. Often navigational or informational about a specific entity; pure organic ranking is harder.

Local pack. The query has local intent. Dedicated local SEO work is required; standard organic optimisation alone won’t compete.

Sitelinks under top result. Google has recognised the top result as authoritative; sitelinks expand its real estate further. Difficult to displace from below.

Ad density. Heavy ad presence above the fold (3-4 ads) reduces available organic clicks substantially even at position 1.

Reading the actual ranking pages

Beyond the SERP features, the pages themselves tell the most useful story. For each top-3 to top-10 result:

• What’s the title and intro? What angle does it take?
• What’s the structure? Look at the H2s as a table of contents.
• Is the answer to the query in the first 200 words, or is it buried?
• What’s missing? Where is the existing content thin, outdated, or wrong?
• Who wrote it? Named author with credentials, anonymous brand content, vendor-promotional?
• What internal links does it have? Pages with strong internal cluster support are harder to displace.
• What backlinks does it have? A quick Ahrefs check shows whether the page has earned its position or coasted on domain authority.

The gap between what’s there and what could be there is the opportunity. If every top result fails to address an obvious sub-question, that’s the angle a new page can win on.

SERP analysis tooling

Manual SERP review is slow but produces the deepest insight. Tools that accelerate the process:

• Ahrefs / Semrush SERP overview. Shows ranking pages with their domain authority and backlink count.
• Surfer SEO, Clearscope, Frase. Content optimisation tools that summarise top-ranking content patterns. Useful for fast SERP characterisation; less useful for deep angle analysis.
• Browser extensions like SEO Minion or Detailed for at-a-glance SERP feature identification.

For high-priority pages, do the manual analysis. For bulk keyword evaluation, use tools to filter; do manual SERP review on the surviving shortlist.

What changed when Google removed the 100-results parameter?

In September 2025 Google stopped supporting the num=100 URL parameter, which had allowed tools and practitioners to pull the top 100 results in a single request. Google confirmed the parameter is not supported.² Three consequences bear directly on SERP analysis.

Deep SERP data became less reliable. Rank trackers now need ten requests where one previously sufficed, and coverage below roughly the top 20 thinned out as a result. One analysis of the change found 77% of sites lost keyword visibility in their reporting.³ When a tool tells you what the whole top 100 looks like, treat the tail of that list with more caution than you would have before.

Search Console data rebased. Automated crawlers appending the parameter had been generating impressions of their own. Once it went, impressions fell and average position appeared to improve for many properties, with no change to the site itself.⁴ Any comparison that straddles September 2025 needs that discontinuity accounted for.

Direct reading of the SERP gained value. The one source that did not degrade is the results page itself, examined manually. The tooling shortcut got worse; the underlying practice this page describes did not.

Common SERP analysis mistakes

Mistake	Effect
Skipping SERP analysis entirely	Producing content that doesn’t match intent
Reading only titles, not actual page content	Missing depth and angle signals
Ignoring SERP features	Overestimating organic CTR
Defaulting to the same content format for every query	Mismatched intent on most queries
Targeting head terms based on volume alone	Ignoring competitive reality visible on the SERP

SERP analysis and AI search

The SERP is increasingly an AI-mediated surface, and reading it means reading two things at once: the ranked results, and the synthesised answer sitting above them.

For a query that returns an AI Overview, record:

• Which sources are cited, and in what order. The cited set frequently differs from the organic top ten. A page can be cited without ranking well, and rank well without being cited.
• Which passage of each cited page was used. Retrieval operates on passages rather than whole pages, so the useful question is not why a page was chosen but which section answered the question, and how that section was structured.
• Whether the Overview finishes the job. If it fully answers the query, organic click-through is suppressed and citation is the realistic goal. If it leaves the task incomplete, the click is still available and the ranked result still matters.

AI Mode has to be checked separately, because it runs its own retrieval rather than reflecting the page you are looking at.

For queries where AI surfaces appear, the optimisation question becomes two-part: how do we rank in the conventional organic results, and how do we earn citation in the synthesised answer? The two are correlated but not identical, and they are measured differently, by position and click-through on one side and citation rate on the other.

Frequently asked questions

How long does a SERP analysis take?
A thorough analysis of a single high-priority query takes 30-60 minutes. Bulk analysis of dozens of queries via tools takes minutes per query but produces shallower insight.

Should I do SERP analysis from incognito or from my normal browser?
Incognito mode (or a private browser session). Personalisation, location, and history all affect what you see. For consistent baselines, also use a tool like Ahrefs SERP Overview which serves desensitised SERP data.

How often do SERPs change?
Major SERPs reshuffle constantly. The dominant content type for a query is more stable; that’s the pattern that matters most for content planning. Re-check the SERPs behind your priority pages quarterly rather than annually: intent classifications shift, and AI surfaces appear on and vanish from individual queries faster than the organic results reorder.

Footnotes

1. AI Mode is now available in more languages and locations around the world — Google ↩

2. Google Search confirms it does not support the results per page parameter — Search Engine Land ↩

3. 77% of sites lost keyword visibility after Google removed num=100 — Search Engine Land ↩

4. Why Google Search Console impressions fell (and why that’s good) — Search Engine Land ↩