A SERP (Search Engine Results Page) is the page that a search engine builds in response to a query. It is not an archive and not a fixed ranking: it is generated on the spot, for that search and for the person doing it, and two people typing the same words at the same moment can see two different pages.
Most people know this part. What almost every definition leaves out is the practical consequence: the SERP you see and the SERP Search Console describes with an "average position" are two different things. Most misunderstandings about rankings start there, and further down you will find the numbers from three sites we manage that show it.
What a SERP looks like in 2026
The ten blue links are a memory. Take "accountant london" searched on Google UK from London (DataForSEO, Google UK, September 2026). Before the first organic result the page shows a row of directories (Bark, Yell, Tipped), a local pack with three firms and their ratings (4.9 from 37 reviews, 5.0 from 106, 5.0 from 58), an AI Overview and four People Also Ask questions. The first classic result, saffery.com, is the seventh element on the page. The mix changes with the type of query.
It is worth no longer thinking of the SERP as a list and starting to think of it as a set of blocks competing for the same vertical space. These are the ones you come across most often.
| Block | What it contains | How you get into it |
|---|---|---|
| Organic results | Title, URL, descriptive snippet | Indexing and ranking, no direct cost |
| Google Ads | Text ads at the top and bottom, Shopping at the top or side | Paid auction on the keyword |
| AI Overview | Generated answer, with a few side links to sources | No application: Google picks the sources itself |
| Featured snippet | Paragraph, list or table extracted from a page | Answering the question in an extractable way, usually already in the top 10 |
| People Also Ask | Related questions in an accordion | Covering the satellite questions of the query |
| Knowledge Panel | Card about an entity: companies, people, places | Being a recognised entity, with consistent data across sources |
| Local pack | Map and three business listings | Google Business Profile listing and proximity signals |
| Images, videos, News | Vertical carousels inside the page | Optimised files, structured data, accredited source for News |
| Sitelinks | Subpages under the main result | Clear architecture and consistent navigation |
The practical point is that these blocks do not add up, they push each other down. If a query triggers an AI Overview, an image carousel and four ads, the first organic result can end up below the fold even when it is in first position. The guide to AI Overviews goes into detail on how Google builds that block, the page on when Google ignores the meta title explains how it chooses the title of your result, and the page on featured snippets explains how to win the space just below.
How often each block appears
The most detailed public tracking of how often each block appears is American. Semrush Sensor, whose data is quoted in the Semrush guide from October 2025, measures the ten most common features in US desktop SERPs as follows.
| Block | Frequency (desktop, United States) |
|---|---|
| Related searches | 95.54% |
| Sitelinks | 77.48% |
| People Also Ask | 67.79% |
| Images | 50.63% |
| Video | 46.65% |
| Video carousel | 26.87% |
| Knowledge Panel | 23.83% |
| Image pack | 23.01% |
| Local pack | 17.62% |
| Popular products | 12.69% |
These are US figures and do not carry over to the UK as they are: Ofcom puts the share of UK searches showing an AI Overview at about 30%, a different mix again. The table works as a checklist, not a forecast. It tells you which blocks to look for when you open the SERP for your keyword, and it says something often missed: related searches, sitelinks and People Also Ask appear almost every time, while the blocks everyone focuses on are much rarer.
Why your SERP is not mine
Google assembles the page taking into account language, country, approximate location, device, account history and, since AI Mode arrived, how it has decided to break the question down. In the official Google Search Central documentation the mechanism is called query fan-out: the engine runs several related searches across subtopics and different sources, then puts the answer together. It means the same question can trigger different search paths in two different sessions.
On the same page Google writes a sentence worth quoting: there are no additional requirements or special optimisations to appear in AI Overviews or AI Mode. The controls are the usual ones, nosnippet, data-nosnippet, max-snippet, and they serve to limit what is shown, not to apply.
In the UK the pool of people who pass through these pages is most of the adult population: according to Ofcom's Online Nation 2025 report, published on 10 December 2025, Google Search is used by 82% of adults and handles 3 billion searches a month. The SERP is where those people decide who deserves a click, and it is not a neutral place.
Average position is an average of different things
Here is the part missing from most guides to the SERP. Search Console reports an average position for each query, but that figure comes from impressions collected on SERPs that differ from one another: some with an AI Overview, some with four ads on top, some on mobile with an image carousel in front. The average of a buried first position and a clean fifth position describes neither of them.
We looked at the raw data of three Italian Search Console properties we manage, for the quarter from 15 May to 15 August 2026. They are three sites of different sectors and sizes, held together only by the fact that the traffic measured is Italian.
| Site (Italian traffic, 15 May - 15 Aug 2026) | Impressions | Clicks | CTR | Average position |
|---|---|---|---|---|
| visilay.com, agency site | 15,287 | 57 | 0.37% | 45.0 |
| B2B site, services for the dental sector | 4,593 | 54 | 1.18% | 26.2 |
| Outdoor equipment e-commerce | 2,319 | 64 | 2.76% | 21.0 |
| Total | 22,199 | 175 | 0.79% | n/a |
The striking number is not the CTR. It is another one: adding up the queries Search Console shows in full for the three sites, 1,646 in total, 1,612 did not produce a single click in three months. 97.9% of the queries for which these sites appeared in the SERP is visibility that never turned into a visit.
The obvious explanation is position: with an average between 21st and 45th you are on page two or three, where nobody goes. But that is exactly the useful information. Appearing in the SERP and being in the SERP the user actually looks at are two different conditions, and a report that says "we are present on 1,646 queries" describes the first while pretending to describe the second. If you want to read a click-through rate without being misled, the page on click-through rate explains it at more length.
What happens to clicks when the SERP answers on its own
The most solid data on this comes from the Pew Research Center, which in March 2025 tracked the real behaviour of 900 US adults across 68,879 Google searches. 18% of searches produced an AI summary. When the summary was there, users clicked a traditional result in 8% of visits; when it was not, in 15%. The link to a source cited inside the summary was clicked in 1% of visits. And 26% of visits with an AI summary ended there, against 16% without.
The sample is American and should be read for what it is, but distribution data confirms the direction. A Semrush study of more than 600,000 keywords in the US desktop database, between November 2025 and April 2026, measured a 71% increase in SERPs with an AI Overview for commercial-intent queries, while for transactional ones the figure fell by 5%. In Finance the growth was 231%, in Computers & Electronics 107.62%. On the most expensive Jobs & Education keywords, more than 60% show an AI Overview, and the average CPC of those keywords is $5.02 against $1.51 for the same categories without an AI Overview.
In plain terms: the generative block is moving exactly to where purchases are decided, the band of queries where the user is comparing options and has not yet chosen. Anyone selling B2B services notices this before anyone selling products through a basket. We have written about what changes in practice in GEO vs SEO and in the guide to Google AI Mode.
Read the SERP before writing the page
The SERP is also the most reliable specification document there is for understanding what Google expects from a page. Before opening the editor, spend half an hour on it, in this order.
- Look at the dominant format. If the first eight results are guides and you were about to build a service page, the service page will not get in. Search intent is set by the SERP, not by your content plan.
- Count the non-organic blocks. Ads, AI Overviews, maps, carousels. If the first blue result sits halfway down the screen, the value of first position on that query is already halved.
- Read the subheadings of the top results. H2s that repeat on three pages out of five are the sections users expect to find.
- Open the People Also Ask questions. They are a free FAQ outline, already written by Google in the vocabulary of the people searching.
- Repeat the search on mobile and from another city. If the SERP changes, you have a query with a local component and you need a different page.
A concrete example of what this reading produces. For Macropix, a Milan-based maker of custom LED screens, the heaviest keyword in Italy is "ledwall": 4,400 searches a month on Google Italy. The SERP showed manufacturers and catalogues, not articles. By working on the product and installation pages instead of the blog, the site went from position 88 to position 2 and now holds a 25.55% share of voice on that group of queries. The full case is in the guide to SEO for manufacturing.
What to measure instead of position
If average position is an average of different SERPs, you need something more stable alongside it. Three metrics hold up better.
- Share of voice on the group of queries that matters. How much of the total visibility available on a set of keywords is yours. It is not thrown off by a query giving you impressions today and none tomorrow.
- Clicks per query, not total clicks. The total rises simply because the number of queries you appear for has grown. The ratio tells you whether SERP presence produces visits or stays as exposure.
- Presence in the blocks, not just in the list. Appearing in a featured snippet, in People Also Ask or among the sources of an AI Overview is worth more than fifth organic position on many queries.
The first two come from Search Console, by crossing query and page data. The third needs a SERP tracking tool, such as Semrush, Ahrefs or Sistrix, or a manual check on a sample of queries always done under the same conditions. When choosing which queries to track, it helps to think in terms of long tail and short tail, because the two behave differently: long queries trigger fewer non-organic blocks and convert better.
How to win space in the SERP
There are no shortcuts, but there is an order of priority. The pages that take up the most space are those structured so that different pieces of the content can be extracted on their own: a crisp definition at the top for the featured snippet, a table with your own data for citation, an FAQ block for related questions, structured data so the machine understands what the page is about.
The rest is ordinary, slow work: keyword research done on the SERP of the market you target rather than translated from another language, correct structured data, and attention to confirmed ranking factors rather than the ones people gossip about. If you would rather have someone else do it, our SEO services and our work on visibility in answer engines start from exactly this reading of the SERP. And if your starting point is further back, the guide to what SEO is lays out the basics.
SERP FAQs
SERP stands for Search Engine Results Page. It is the page that Google, Bing or another search engine generates in response to a query.
Organic results are selected by the algorithm and cannot be bought: you earn them through SEO work. Ads are spaces bought through the Google Ads auction and carry the Sponsored label. They occupy different positions on the same page and can sit alongside an AI Overview.
Because your own search is personalised by history, account and location, while Search Console reports an average of impressions collected across different users and SERPs. For a cleaner check, use a private window and still compare it with the aggregated data, not with your own impression.
There is no single figure valid for every sector. Ofcom puts it at about 30% of UK searches in its Online Nation 2025 report. The Pew Research Center found an AI summary in 18% of 68,879 Google searches tracked in March 2025 in the United States, and Semrush measured a 71% increase in commercial SERPs with an AI Overview between November 2025 and April 2026. The share varies a lot by sector and query intent.
No. In the Search Central documentation Google states that there are no additional requirements or special optimisations to appear in AI Overviews or AI Mode. There are only controls to limit what is shown, such as nosnippet, data-nosnippet and max-snippet.
Yes, in the mix of blocks and in how sources are selected. In the UK Bing handles a much smaller share of searches than Google, but its index also feeds some answers from conversational assistants, so appearing there has a value that does not show up in direct traffic data.
Something rarely said in the industry: the SERP is becoming a reading interface for machines before people. An AI Overview, ChatGPT or Perplexity do not read the page as a reader does, they lift pieces from it. From here on the useful question is not only what position you are in, but which fragments of your page are self-contained enough to be taken and cited in full. It is a different way of writing, and whoever gets there first has an advantage that does not depend on the number of inbound links. Zero-click searches tell the other side of the same story.
