SERP analysis for link building: how to read a results page

A SERP, short for search engine results page, is the page a search engine returns for one query. It holds more than a list of ten links. A results page in 2026 can carry ads, an AI Overview, a map with local businesses, a featured snippet, a box of related questions, images, videos and the ordinary text results, and each of those blocks is filled by a different system with different inputs.

For link building this matters because the results page is the cheapest evidence available about what a keyword will cost. It shows what kind of page Google has decided answers the query, which sites hold the positions, and how much of the screen is open to an ordinary web page at all. Reading it carefully takes a few minutes and should come before any link data, because the link numbers only mean something once you know which results are worth comparing yourself with.

What is on a results page today

Google’s own reference for this is the visual elements gallery in Search Central, last updated in February 2026. It calls the classic listing a text result, “formerly known as a web result or plain blue link”, made up of attribution, a title link and a snippet, sometimes with sitelinks or rich attributes such as review stars. The gallery also notes that a text result “may display differently depending on a variety of factors”, including the device, the query and the language. Around the text results sit the other blocks, and the table below sets out the main ones by what fills them.

Block	What it shows	What decides who appears
Text results	Pages from the web index, with title link and snippet	Google’s ranking systems, where links are one input among many
Ads	Paid listings from Google Ads	Ad Rank, which Google Ads help says is calculated from the bid, ad and landing page quality and other auction factors
AI Overview	A generated summary with links to supporting pages	Google’s AI features, which its documentation says may use a query fan-out technique across related searches
Local results	A map and a short list of businesses	Relevance, distance and prominence, according to Google Business Profile help
Featured snippet	A box that shows the descriptive snippet first, above the link	Google’s systems decide whether a page “would make a good featured snippet”
Related questions	A group of questions with expandable answers, known as People also ask	Chosen by Google, and described in its gallery as a cluster of questions related to the initial search

The third column is the one to study. Ads are bought, and no number of links moves a page into them. Local results draw on a business profile and on prominence, which Google’s help page says is based on information such as “how many websites link to your business and how many reviews you have”. A featured snippet is an ordinary result that Google has chosen to lift into a box, because its documentation says its systems decide whether a page would make a good one and then promote it, so the snippet rewards a page that is already competing in the text results. That leaves the text results as the place where a link campaign has its most direct effect, and the first job is to see how much room they have.

Why the page comes before the link data

Most guides start with a backlink tool. The site’s lesson on how many backlinks you need gives a seven-step estimate built from the referring domains of the pages that rank, and that estimate is only as good as the list of pages fed into it. Its second step, removing results that are not like you, is a judgement about the results page, and it is the step people rush. If the top ten holds a national newspaper, a marketplace and an encyclopaedia, their link counts describe businesses of a different kind, and a median taken across them sets a target no small site can reach or needs to.

There is a second reason. Google’s Gary Illyes said in April 2024 that Google needs very few links to rank pages, and in Ahrefs’ study of one million keywords, published in January 2025, referring domains correlated with rankings at 0.255. The same study found the correlation stronger for local searches, at 0.33, and weaker for commercial and transactional ones. A correlation of that size means links explain part of the order on the page and much of it comes from elsewhere: how well the page matches the query, the standing of the site, and how searchers respond. The results page shows those other factors directly, so reading it first protects you from buying links to fix a problem that links do not cause.

How to read the page for a keyword

What follows is a working method. No study has measured whether it predicts link needs better than any other routine, and its value is that it forces a set of observations into a fixed order before money is committed. Search the keyword in a private window, in the country you are targeting, on both a phone and a desktop, and work through the page from top to bottom.

1. Note every block above the first text result: ads, AI Overview, local results, featured snippet. Count how far down the first text result sits.
2. Classify the query from what you see. A map means local intent, product listings mean buying intent, long guides mean a question to be answered.
3. Record the page type of each of the top ten text results: homepage, category page, product page, guide, list article, tool, forum thread.
4. Record the site type of each: brand, retailer, publisher, review site, directory, government or university, community.
5. Mark the results that resemble your own page and site. These are your real comparison set.
6. For the marked results only, look up referring domains to the page and to the whole site.
7. Open the links behind two or three of those figures and note where they come from.
8. Write one sentence on what a page like yours would need to appear here.

Step three often settles the matter before any link is counted. Suppose a company that makes payroll software for restaurants wants to rank its product page for a broad phrase about restaurant payroll. If nine of the ten text results are guides and comparison articles from publishers, the results page is saying that Google reads the query as a request for advice. Links pointed at a product page will struggle against that reading. The better plan is a guide of the company’s own, or a place in the comparison articles that already rank, which is the approach covered in best-of listicles. Which pages need links covers the related choice of link target within a site.

Judging how many links and what kind

Steps six and seven turn the reading into numbers, and the distinction between page-level and site-level figures does most of the work. Use referring domains and not raw link counts, for reasons that lesson explains. Three patterns come up again and again. In the first, the comparable pages each have a meaningful number of referring domains of their own. That is a page-level contest, and the gap between their median and your figure is a fair first estimate of the work, to be refined with a link gap analysis.

In the second pattern, the ranking pages have few or no links of their own but sit on sites with very large profiles. Here the page is carried by the site, and pointing a handful of links at your equivalent page will not reproduce that. The honest conclusion is that the keyword needs site-wide authority, which takes longer and costs more, or that a narrower keyword with weaker sites in the results is the better target. In the third pattern the results are mixed: one or two small sites with modest link profiles sit among the large ones. That is the most encouraging sign a results page can give, because it shows that a site of your size is already being allowed to rank, and those small sites are the ones whose links are worth opening one by one.

What kind of links to pursue comes from step seven. If the comparable pages are linked from trade publications, the route is editorial coverage and something worth citing, such as a linkable asset. If their links are mostly from directories and local organisations, the query behaves as a local one. In that case reviews and local citations will matter alongside links, which fits Google’s own description of prominence. If the links are plainly bought placements on general blogs, you have learned what the market does without learning that it works, since Google may be ignoring those links and ranking the page for other reasons. How to check backlinks covers the mechanics of pulling these lists, and the cost calculator turns a referring domain gap into a rough budget.

AI Overviews and the space above the results

The blocks above the text results change what a ranking is worth, and the reading should record them even though links cannot buy them. Google’s help page for searchers describes an AI Overview as an AI-generated snapshot with links to dig deeper, shown when its systems judge that generative AI can be especially helpful. For site owners, Google’s documentation on AI features, updated in December 2025, says there are “no additional requirements to appear in AI Overviews or AI Mode”, and that these features may use a query fan-out technique, issuing multiple related searches across subtopics. The practical meaning is that the pages cited in an overview are not always the pages ranking beneath it.

The figures on the site bear that out. Ahrefs analysed 863,000 results pages in March 2026 and found that 37.9% of the URLs cited in AI Overviews ranked in the top 10 for the same query, with the rest further down or beyond position 100. So when an overview appears for your keyword, note which sites it cites and whether they match the text results. If they differ, there are two contests on one page, and how AI search chooses sources explains what is known about the second one. For the link estimate, the overview mainly affects value: a first-place text result that sits below an overview, three ads and a map will be seen by fewer people than the same position on a plain page, which belongs in the decision about whether the keyword is worth the links at all.

Limits of the method

A results page is a snapshot. It changes with location, device and time, and a single search on one afternoon can catch a reshuffle after an update. Run the same reading again a few weeks later for any keyword that will carry real budget, and prefer a rank tracking or backlink tool that stores the history of the results, so that you can see whether the sites you marked as comparable have held their places or only just arrived.

The link figures have their own weakness. They come from a third-party crawl that misses some links and counts others that Google discounts, so two tools will give two different medians for the same ten pages. And the method cannot see causes. A small site ranking among large ones may be there because of its links, its content, its brand, or a quirk that will not last. The reading narrows the question and gives a range, which is what a budget needs. It should be written down with the date and the country, kept beside the link gap figures, and revisited when the page has been live long enough for the first links to be counted.