SEO Statistics for 2026 (With Sources, and Where They Disagree)

Most SEO statistics pages quote a number, attribute it to nobody, and repeat a figure that has been copied between blogs since 2019. The click-through rate for position one is the worst offender: you can find it confidently stated as 27.6%, 39.8% and 31.7% on pages published the same month.

Those numbers come from different studies measuring different things, and all three can be honest. This page gives you the figures with the study named next to each one, and flags the places the major sources contradict each other rather than picking whichever is most quotable.

How this page was built, and what changed

This page previously listed over 110 statistics. Most of them had no source attached, which made them impossible to verify and impossible to defend if a reader asked.

We cut it to the figures we could attribute to a named, findable study. That is fewer numbers and considerably more useful ones. Where a figure comes from a vendor with a commercial interest in the result, which is most of this industry, that is noted rather than hidden.

One warning that applies to everything below. Almost all SEO statistics come from vendor-run studies of vendor-accessible data. They are the best available evidence and they are not neutral. Treat them as directional.

The headline shift: clicks are disappearing

This is the most important trend in the data and the one that should change how you pick keywords. The transactional-versus-informational split is the actionable part: commercial queries still produce clicks at more than twice the rate of informational ones. Which queries survive that shift is a question about meaning rather than strings, which we cover in semantic search and SEO.

AI Overviews

Where the sources disagree: AI Overview prevalence is reported anywhere from 20% to 48% of searches depending on who is measuring and which keyword set they track. Those are not reconcilable, and the gap is mostly explained by keyword sampling. Anyone quoting a single confident number for this is over-claiming.

Click-through rate by position

This is the most contested area in SEO data, so here are the competing figures side by side rather than a single number. Ahrefs produces several of the studies quoted here, and how it compares with the optimisation tools is covered in Surfer SEO vs Ahrefs.

Why they disagree. Studies differ on whether they include branded queries, which push position-one CTR far higher, and on whether they measure across all query types or only informational ones. Use these as a shape rather than a forecast, and use your own Search Console data for anything you plan to act on, because it reports your actual CTR rather than an average of someone else's.

Organic search as a channel

That second line matters more than the first. The 53% figure is the most-copied statistic in SEO and it describes an average across wildly different businesses. Your own analytics is the only number worth planning against. B2B sites sit furthest from that average, because the buying cycle is long and the useful queries are small, which is also why we keep a separate shortlist of the best B2B SEO tools.

AI search referrals

The newest category in this data, and the one changing fastest. Providers claiming to move it are covered in our explainer on AI SEO services.

Read this section sceptically. AI referral volumes are still small in absolute terms for most sites, which makes conversion-rate comparisons volatile. A channel that swung from 38% worse to 42% better in a year is a channel with unstable measurement, not a settled finding. The direction is real; the precision is not.

LinkedIn as a parasite SEO surface

This is the one section on this page we ran ourselves. Everything else here is somebody else's study. We collected this one in August 2026 because nobody had published control-group data on it, and we sell SEO software, so read it with the same suspicion you would apply to any vendor study. The method and the raw limitations are at the end of the section.

The collection: 200 hand-built SEO and B2B SaaS queries, each pulled with the top 20 organic results and the full AI Overview source list. The same 200 queries sent word for word to ChatGPT, Claude and Perplexity with web search on. Then 400 further searches to build a control group of LinkedIn content Google has indexed but does not rank. US desktop, English.

Where LinkedIn shows up

That last row is the one worth keeping. Being cited by an AI Overview was 11 times more likely when the page already ranked, which makes AI citation largely downstream of classical ranking rather than a separate discipline.

The engines cite very differently

This is a retrieval-depth difference, not a model-quality one. An engine that pulls 19 sources per answer reaches into social platforms to fill them. An engine that pulls 2 goes to the brand's own site. Treating "AI visibility" as one channel hides that split.

Which queries LinkedIn wins

Best-X is the category most people write LinkedIn posts for and the one LinkedIn loses. Across 40 such queries LinkedIn appeared in 4 and was cited by a single AI Overview.

What separates ranking LinkedIn content from the rest

These come from the control group: 879 LinkedIn URLs with usable text, 106 that rank in Google's top 20 and 773 that Google has indexed on the same topics but does not rank.

The useful negative result. How the content is written did not predict whether it ranked. Length, heading count, paragraph structure, hook style and list usage all failed inside both formats. What predicted ranking was query match in the headline, a single backlink, engagement and age.

Method and limitations

The 200 queries were written across 8 intent classes before any data was collected, so the intent splits above are not selected after the fact. The control group is LinkedIn content Google has indexed and returns for a site: query on the same topic, so content Google never indexed is absent and the rank rates compare buckets against each other rather than giving absolute probabilities.

Reaction counts come from Google's own SERP breadcrumb and are bucketed at "200+", so they are coarse. Everything is observational: the backlink result survives stratifying by both age and engagement, which removes the two obvious confounds, but a link and a ranking could still share an unmeasured cause.

Chatbot prompts were sent in Google-query phrasing rather than conversational phrasing, which keeps the engines directly comparable but is not how people talk to a chatbot. Gemini was excluded because its citation URLs resolve through Vertex AI redirects that hide the destination domain. 51 ranking Pulse articles and 55 ranking feed posts is a small numerator, so the null results on structure are better read as no effect large enough to detect at this sample size than as proof of no effect.

Every query, result and citation is in the raw dataset behind this study. If you want to check a specific number, ask and we will send the CSVs.

Local search

Local is where the data most consistently supports a specific action: complete the profile, post, and collect reviews continuously. Our breakdowns of SEO for dentists and SEO services for restaurants cover what that looks like in practice.

What these numbers should change

Stop targeting purely informational head terms. At 74% zero-click on informational queries and 83% when an AI Overview appears, ranking first on "what is X" produces impressions and very little else. Transactional and commercial queries still click through at roughly twice the rate. Producing the commercial pages that remain worth targeting is what our roundup of automated SEO software is about.

Stop quoting position-one CTR to clients. The honest version is that it depends on the query type, whether the brand is yours, and whether an AI Overview is present, and that credible studies disagree by more than ten percentage points. What a position actually means is unpacked in our guide to SEO positioning. The wider job of running this for a business, including what to report and how often, is covered in what SEO management involves.

Start measuring AI referrals separately. They are small, they convert well, and they are invisible in most default analytics setups. Whatever the true numbers are, you cannot manage a channel you are not counting.

Use your own data for anything you act on. Every figure on this page is an average across sites that are not yours. Search Console reports your real positions, your real CTR and your real query mix, and it is free. Our free SEO ROI calculator does the same with your own numbers. If you are doing the work yourself rather than briefing an agency, how to do SEO optimization yourself is the step by step version.

The statistics we removed and why

The previous version of this page carried figures such as blanket claims about content length and rankings, precise ROI multiples for SEO as a channel, and several conversion benchmarks with no study behind them.

They came from sources that either could not be traced or turned out to be one blog citing another blog citing a 2019 vendor post. A statistic that cannot be traced to a study is not evidence, and on a page that other people cite, repeating one propagates it.

If you are compiling your own statistics page, the test we applied is simple: can you name the organisation that produced the number and the year they produced it. If not, cut it.

A note on citing this page

Every figure above names its source. If you cite something from here, cite the original study rather than us, because we did not run most of these and the primary source is what your reader needs. The exception is the LinkedIn section, which is our own first-party study: cite that one as Distribb, August 2026, and note that we sell SEO software.

Where we have written "aggregated industry data", that means the figure appears consistently across several 2026 compilations but we could not trace it to a single primary study. Treat those rows as weaker than the named ones.

If the data has changed your plan

The clear implication of the zero-click numbers is that fewer queries are worth targeting than in 2020, and the ones that remain are more competitive and more commercial. That means fewer, better pages against queries where the click still happens. For a portfolio or an in-house team rather than a single site, use the enterprise SEO ROI calculator.

Distribb runs keyword research, writing, publishing and internal linking on a schedule, with a backlink exchange for the off-page half, which is the production side of that shift. There is a 3-day free trial. It will not tell you which queries survive the zero-click problem, so read the transactional-versus-informational split above before pointing anything at a keyword list. If you would rather buy that as a service, we ranked the best AI SEO services. How often that schedule should fire is its own question, answered with data in how often you should blog for SEO.