Your website is not only for people browsing, it's read by search crawlers, assistive technology, and a growing set of AI agents that summarise, cite, or train on what they find. The mistake we keep seeing on client estates is treating those as separate projects: an SEO pass in marketing, an accessibility ticket before go-live, and "AI readiness" as a blog post nobody owns.
The surface is the same and just needs to be taken care of: rendered HTML that is structured, visible, and intentional.
People, search, and agents
Accessibility has a clear bar. WCAG 2.2 Level AA is the working standard most UK buyers ask for. For public-sector and education bodies, the Public Sector Bodies (Websites and Mobile Applications) Accessibility Regulations 2018 turn that into a legal conformity story.
Technical SEO still rests on Google's basics: crawlable URLs, honest titles and canonicals, useful structured data, and pages that actually render the content your users care about. Google Search Central is still the source of truth for how crawlers interpret that. Social previews are fundamental too - Open Graph and card meta decide what appears when someone pastes a link into LinkedIn, Slack, or Messages. That footprint is part of SEO delivery, and your digital footprint.
Agent readability is newer, but the pattern is familiar. Named bots in robots.txt (training versus citation versus user-fetch), optional briefing files such as llms.txt, and enough server-rendered text that a fetcher does not hit an empty shell. None of that replaces search rankings or WCAG conformance. But it does decide whether answer engines and assistants can quote the pages you want them to quote - and whether you meant to allow training crawlers at all.
The real SEO problem
Across member platforms, education schemes, and product sites we audit - including SCONUL Access under SEO review and accessibility compliance - the failures rarely look exotic. In fact all they need is a knowledgeable and experienced eye to monitor the digital landscape and ensure brand consistency across the board:
Contrast that only fails in one theme - Dark mode passes; light mode fails on shared chrome - buttons, consent links, status badges - because tokens were never checked as a matrix.
Strong page chrome, weak machine signals - Titles and schema look fine on a few hub pages, then thin out across the sitemap. Or the reverse: SEO hygiene is clean, but every AI bot inherits a blanket User-agent: * with no explicit training-versus-citation policy.
Missing or decorative llms.txt - Either absent, or a dump of every URL instead of a curated briefing to services, proof, and contact.
SSR gaps - Content that only appears after client JavaScript runs is invisible to many crawlers and fetchers - and often harder for assistive tech when names and structure are late too.
Scores sold as certificates - Automation (axe-core, rendered SEO checks, Lighthouse) is excellent for readiness. It typically catches a large share of issues - not all of them. Keyboard paths, focus order, media alternatives, and Search Console coverage still need guided and manual work.
How we solve your SEO at Pendulum
We treat SEO and accessibility as a paired audit, then remediate with the same team that designs and builds.
Rendered audits, not source-only guesses - Pages are collected in a real browser so titles, meta, schema, contrast, names, and landmarks match what users and bots actually get.
Defined catalogs and honest scores - Technical SEO (including social footprints and a separate agent-readiness view) and WCAG 2.2 AA checks produce developer detail plus plain-language client summaries. Agent policy does not get smuggled into an SEO score to make a dashboard look greener.
Theme and viewport matrices - Light/dark (or high-contrast) themes are first-class - shared tokens that fail in noon mode show up before a public-sector review does.
Policy before fashion - For agents we ask a product question first: allow citation, restrict training, or the reverse? Then we write named User-agent rules and, where it helps, a short llms.txt that points at the pages you want quoted.
Fix and revisit - Findings become a prioritised backlog. For regulated estates we bake in periodic reviews so conformity does not rot between releases.