Socially acceptable tech masochism club

The Web Almanac returns for another year, slightly later than usual but packed full of as interesting data as ever! They followed a similar process to previous years with teams of dedicated experts split into a number of "chapters" with each of them pouring over the terabytes of HTTP Archive data to pull out interesting stats on the current state of the web for a particular topic. As always the data—and the SQL queries used to find them—are made available for even more digging if the hundreds of thousands of data points in the Web Almanac itself only piques your interest more.

This is a very fancy way of saying volunteer subject matter experts put themselves through crucibles analyzing 244 TB of data across 16.2 million sites to let the denizens of the web know what's happening, changing, and absolutely confounding.

Questions like "How big is a typical web page?", "What are the most popular CMS platforms?", or "How much unused JavaScript is on the web?" are answered with real data in the Web Almanac.

To celebrate I asked contributors to give me their best snarky takes:

"The 2025 Web Almanac SEO chapter shows a web that’s broadly holding together, but not always thriving on its own merits. Many core fundamentals look stable on the surface, yet a lot of progress is coming from CMS defaults and plugins papering over cracks rather than "deliberate" optimisation. There’s still plenty broken under the surface: old bot references, inconsistent markup, and whole classes of SEO work that remain invisible unless you’re pressed up against it. But the context has shifted. robots.txt is shifting intent toward AI concerns, AI crawlers are being explicitly managed, and experiments like llms.txt suggest where things might head next, even if adoption is still tentative or largely accidental. SEO isn't “done”, it’s that most has moved because platforms moved it, even thought SEOs are heroically working away. As AI systems increasingly decide what gets understood and cited, the unglamorous, hard-to-see work still matters, even if you can't see it from orbit."

Chris Green, SEO chapter

"The web is still doing a very hungry caterpillar impression and growing heavier and denser, but there's still hope and tools to slim it down. Being able to see the impact of weight on real users' core web vitals this year was super interesting. One of the most surprising things was INP was much less affected by page weight than I thought it would be. Still correlation, but I thought the effect would have been more dramatic."

Dave Smart, Page Weight chapter

"Wasting resources is of course the in thing with LLM's but javascript heavy sites and throwing whatever images against the wall to see what sticks is absolutely not helping. Money is being left on the table because not enough sites are checking their own homework"

Richard Barrett, Page Weight chapter

"The 2025 accessibility chapter paints a picture of slow but real progress: automated scores are inching up, but core barriers like color contrast, headings, and alt text remain stubbornly common. Regulation and AI are reshaping the landscape without yet delivering a step‑change in real‑world accessibility.

Around 98% of sites now have a `<title>` element, but many titles are still "insufficiently descriptive," which undermines usability and scannability when users switch tabs or rely on screen readers. - About 86% of sites correctly declare a primary `lang` attribute. This supports more accurate automatic translation and correct pronunciation in screen readers, which can indirectly improve experience for international users coming from search. - Roughly 50% of images still have empty or very short alt text, and about 8.5% simply reuse filenames like `image.jpg` as alt text, which wastes a chance to provide descriptive, keyword-relevant context."

Bogdan Lazar & Mike Gifford, Accessibility chapter

This year also sees a new chapter introduced: Generative AI. Christian Liebel, Yash Vekaria, and Jonathan Pagel used not only the dataset of HTTP Archive, but also npm download statistics, and data provided by Chrome Platform Status and other researchers to provide a technology overview, analyze discoverability, and delve into AI fingerprints. My request for a quote was met with silence on slack. I'd like to think the crew finished their chapters and threw their laptops into the nearest body a water to celebrate the start of their well earned rest.

Got FOMO? Good. The Web Almanac is a volunteer project. When the year’s project is announced, that’s your time to look through open issues in the Github repo and raise your hand for where you’d like to contribute. Each chapter has multiple roles that need to be filled. See you in the 2026 edition.

[Read the 2025 Web Almanac// YMMV]