Technical SEO statistics for 2026, explained in plain words.
How fast websites really are, what slows them down, how Google and AI tools crawl them, and how many sites use schema, canonicals and llms.txt. Every technical word is explained, and all sources are listed at the end.
The 6 numbers that matter most.
The short version. Terms like LCP and TTFB are explained in the next section.
Websites that pass all 3 Core Web Vitals on mobile. On desktop it is 60.04%.
Websites with a good server response time (TTFB) on mobile. It is the weakest speed measure on the web.
Share of desktop pages where the biggest item on screen (the LCP element) is an image. On mobile it is 76%.
Major AI crawlers, other than Google’s, that run JavaScript on the pages they visit.
ChatGPT crawler visits that hit a missing (404) page. For Googlebot it is 8.22%.
Websites with a working llms.txt file. 39.6% of those were created automatically by a plugin.
The technical terms, explained.
You will see these words on this page. Here is what each one means.
3 checks Google uses to judge how a page feels to use: loading speed (LCP), how fast it reacts (INP) and how stable the layout is (CLS).
Largest Contentful Paint. How long the biggest image or text block takes to appear. Good is 2.5 seconds or less.
Interaction to Next Paint. How quickly the page reacts when you tap or click. Good is 200 milliseconds or less.
Cumulative Layout Shift. How much the page jumps around while loading. Good is 0.1 or less.
Time to First Byte. How long the server takes to start sending the page. Good is 0.8 seconds or less.
A program that visits web pages to read them, like Googlebot or the ChatGPT crawler.
A tag that tells search engines which version of a page is the main one.
Code that describes your page to machines, like your business name, address and opening hours.
Core Web Vitals in 2026, by device.
This is real data from Chrome users who agreed to share it, measured for whole websites (origins).
| Share of websites that are “good” | Mobile | Desktop |
|---|---|---|
| Pass all 3 Core Web Vitals | 53.21% | 60.04% |
| Good LCP (main content loads fast) | 65.78% | 78.79% |
| Good INP (reacts quickly) | 80.39% | 97.79% |
| Good CLS (layout stays still) | 83.04% | 73.76% |
| Good FCP (first content appears fast) | 58.48% | 75.62% |
| Good TTFB (server responds fast) | 46.00% | 58.47% |
Two things stand out. CLS is the only check where phones do better than computers. And TTFB is the weakest measure of all: fewer than half of websites respond fast enough on mobile. Slow servers and hosting are the biggest speed problem on the web.
Websites passing all 3 Core Web Vitals on mobile
By year, July each year.
The web is getting faster. The mobile pass rate nearly doubled in 5 years. On desktop it rose from 40.53% to 60.04%.
How heavy web pages are.
HTTP Archive tested 15.3 million mobile and 11.3 million desktop pages in August 2026.
| Typical (median) page, August 2026 | Mobile | Desktop |
|---|---|---|
| Total page size | 2,627.9 KB | 3,015.5 KB |
| Images | 881.7 KB | 1,056.1 KB |
| JavaScript | 683.5 KB | 766.6 KB |
| Files loaded per page | 70 | 76 |
| JavaScript files per page | 21 | 23 |
The typical page stays about the same size, but the heaviest pages keep growing. Images are still the biggest part of most pages. Newer, faster connection types are spreading too: HTTP/3 grew from 35.79% to 44.94% of mobile requests in a year.
How Google and AI tools crawl websites.
Vercel measured crawler visits across its network of websites.
| Crawler behaviour | Googlebot | ChatGPT | Claude |
|---|---|---|---|
| Visits a month | 4.5 billion | 569 million | 370 million |
| Visits that hit a missing (404) page | 8.22% | 34.82% | 34.16% |
| Visits that followed a redirect | 1.49% | 14.36% | not given |
Over a third of AI crawler visits hit dead pages. Cleaning up broken links and setting up proper redirects now helps with AI tools as well as Google.
Share of AI bot crawling done to train AI models. Crawling for a user’s live question is under 5%.
How many more pages Googlebot saw than ClaudeBot. Against PerplexityBot it was 167x.
Google also runs JavaScript on every page it crawls. Vercel and MERJ found Google rendered 100% of pages tested. The typical wait between crawling and rendering was 10 seconds, but some pages waited hours. AI crawlers from ChatGPT and Claude download some JavaScript files but do not run them.
How much content needs JavaScript.
98.9% of websites use JavaScript (W3Techs). The question is how much text only appears after it runs.
| Typical home page | Desktop | Mobile |
|---|---|---|
| Visible words after JavaScript runs | 413 | 378 |
| Visible words in the raw HTML | 340 | 316 |
| Extra words that need JavaScript | about 18% | about 16% |
Robots.txt and llms.txt.
Robots.txt is a small file that tells crawlers which pages they may visit. The Web Almanac checked 16.2 million websites.
What robots.txt files look like
Share of websites, July 2025 crawl.
Few sites name AI crawlers in robots.txt: GPTBot 4.5%, ClaudeBot 3.6%, Google-Extended 3.4% and PerplexityBot 2.8%. Cloudflare found only about 37% of top domains have a robots.txt at all, and just 7.8% of those block GPTBot.
llms.txt is a newer file that lists a site’s key pages for AI tools. Only 2.13% of desktop sites have a working one, and 39.6% of those were made automatically by the All in One SEO plugin. It is optional. It will not fix weak pages.
Indexing and canonical tags.
| What pages have | Desktop | Mobile |
|---|---|---|
| A canonical tag | 68% | 67% |
| A canonical pointing to another page | 7% | 9% |
| A noindex tag (asks Google not to index) | 3.5% | 2.4% |
| Broken HTML in the page head | 10.1% | 10.3% |
About 1 page in 10 has broken HTML in its head section. This can stop Google reading your canonical or robots tag. The usual cause is an image, div or link placed in the head by mistake.
SEO plugins set a lot of this automatically. Yoast writes the robots tags on 15.96% of desktop pages, RankMath on 3.56% and All in One SEO on 2.96%. Check your plugin settings once, and it fixes many pages at the same time.
Schema and structured data.
| Type of structured data | Home pages (Web Almanac) | All websites (W3Techs) |
|---|---|---|
| JSON-LD (the format Google prefers) | 43% | 55.5% |
| Microdata | 17% | 21.3% |
| Open Graph (for social sharing) | not given | 72.1% |
| No structured data found | not given | 20.2% |
About 50% of home pages now have some structured data. The most common types are WebSite 37%, Organization 26.74% and WebPage 25%. But 24% of pages use a type that search engines do not recognise, so their schema does nothing.
On-page basics across the web.
Pages with a meta description. Titles are on 98.6% of pages.
Pages with an H1 heading. Only 66% have one with text in it.
Images with an alt attribute on a typical page. 15.38% have blank alt text.
Mobile pages with a viewport tag, which makes them fit phone screens.
Does speed affect sales?
Rise in retail conversions from a 0.1 second faster mobile site, across 37 brands. Average order value rose 9.2%.
Increase in sales when Vodafone made its LCP 31% faster, in a real 50/50 test.
Google says Core Web Vitals are used by its ranking systems, but that “there is no single signal” (Google Search Central). Speed helps, but it will not make up for weak content or a poor profile.
What to do with these numbers.
6 technical fixes, in the order most sites should do them.
1. Check your server speed first
Only 46% of sites have a good TTFB on mobile. Better hosting or caching also speeds up LCP and FCP.
2. Treat slow LCP as an image problem
Images are the main LCP item on 76% of mobile pages. Resize and compress them, and load the main one first.
3. Fix broken HTML in your page head
1 in 10 pages has it, and it can hide your canonical and robots tags from Google.
4. Put key text and schema in plain HTML
AI crawlers do not run JavaScript. Keep services, prices, address and hours in the page code itself.
5. Add LocalBusiness schema
Only about half of home pages have structured data. Use our free schema generator and match it to your Google Business Profile.
6. Clean up 404s and redirects
A third of AI crawler visits hit dead pages. Fix broken links and redirect old pages to the right new ones.
Where every figure came from.
The research behind every number on this page. Each source shows who ran it, when, and how big it was.
Chrome UX Report
Chrome UX Report (HTTP Archive), July 2026 release. Real-user data from Chrome.
HTTP Archive
August 2026 crawl, 15.3 million mobile and 11.3 million desktop pages. HTTP Archive Web Almanac 2025 SEO chapter, 16.2 million websites. Web Almanac 2025 performance chapter.
Vercel and MERJ
Vercel and MERJ, how Google handles JavaScript, July 2024, 100,000+ Googlebot visits. Vercel and MERJ AI crawler study, December 2024.
Cloudflare
Cloudflare AI crawler purpose, August 2025. Cloudflare crawler comparison, January 2026. Cloudflare robots.txt analysis, June 2025.
Deloitte for Google
Deloitte for Google, Milliseconds Make Millions, March 2020. 37 brands.
Google’s Vodafone case study, March 2021. Google Search Central, page experience documentation.
Last checked September 2026. When a source publishes newer data, we update the figure and the date.
Technical SEO statistics, answered.
What share of websites pass Core Web Vitals in 2026?
In the July 2026 Chrome UX Report, 53.21% of websites passed all 3 Core Web Vitals on mobile and 60.04% on desktop. The mobile pass rate was 28.54% in 2021.
What are Core Web Vitals?
They are 3 checks Google uses to measure page experience: LCP (how fast the main content loads), INP (how fast the page reacts) and CLS (how much the layout moves).
What is the weakest speed measure on the web?
TTFB, the time the server takes to start sending the page. Only 46.00% of websites have a good TTFB on mobile.
What slows down LCP most often?
Images. The Web Almanac found images are the LCP element on 85.3% of desktop pages and 76% of mobile pages.
Do AI crawlers run JavaScript?
The major AI crawlers other than Google’s do not run JavaScript. Google renders all pages it crawls. So text that only appears after JavaScript runs may be invisible to AI tools.
How many websites use llms.txt?
Only 2.13% of desktop sites have a working llms.txt file, and 39.6% of those were created automatically by a plugin.
How many websites use schema markup?
About half of home pages have some structured data. JSON-LD is on 43% of home pages in the Web Almanac and 55.5% of websites in W3Techs data.
How often do AI crawlers hit broken pages?
Vercel found 34.82% of ChatGPT crawler visits and 34.16% of Claude visits hit 404 pages, against 8.22% for Googlebot.
Does page speed affect sales?
Yes. Deloitte found a 0.1 second faster mobile site lifted retail conversions 8.4%. In a Vodafone A/B test, 31% faster LCP brought 8% more sales.
Other statistics pages in this series.
Same rules on every page: a date, a source and a plain explanation for each number.
Fix the basics, then check where you rank.
Create your business schema and see where you show up on Google in your area.
Both free to use