AI search: fewer than 10% of sites optimised, Abondance reports

Abondance has published a look at a Fasterize barometer that scores websites out of 100 across five pillars: Search, AI Search, Experience, Accessibility and AI agents. The source reports that fewer than 10% of sites are optimised for AI-driven search. A score above 80 is treated as well optimised. The report does not publish the sample size, the scoring weights or the breakdown by pillar, so the headline figure should be read as a directional signal rather than a benchmark.

What the barometer actually measures

The structure is the interesting part. Splitting "Search" from "AI Search" implies two distinct readings of the same underlying site, even though Abondance frames them as sharing a common foundation. Adding Experience, Accessibility and AI agents as separate pillars pushes the audit beyond classic crawlability and content questions into rendering, usability and machine-readability by agents. Copying from the report, the detail behind each pillar is not available, so teams cannot yet compare their own scores pillar by pillar against a published methodology.

Why the gap matters for acquisition

If the under-10% figure holds broadly, the constraint on visibility in AI-generated answers is less about a new discipline than about fundamentals most sites have not finished. Retrieval-based answer engines depend on content that is crawlable, well-structured, fast to render and clear about entities and facts. That is largely the same groundwork that helps conventional search. The difference is the margin for error: a page that renders poorly or buries its answer may still rank somewhere, but it is far less likely to be quoted.

For teams, the practical read is that AI search is probably not a separate project to staff. It is an extension of technical and editorial hygiene, measured with one more lens.

What a team can check this week

A short, non-exhaustive starting list:

• Confirm key pages render fully without client-side scripts blocking the main content.
• Check that each important page states its answer in plain language near the top, not three scrolls down.
• Review structured data and internal linking on the pages that carry commercial intent.
• Test how headings and paragraphs read when extracted out of context, as an answer engine would.
• Note accessibility issues flagged by standard audits; they often overlap with machine-readability problems.

None of this requires a new tool budget. It requires someone owning the audit and a habit of re-running it.

The open questions

Abondance does not detail how the AI Search score is computed, which signals it rewards, or how the five pillars are weighted. Until the methodology is public, treat the percentage as a prompt to audit your own site rather than as proof of an industry-wide failure. What is clear is that the scoring conversation has moved from keywords to the full rendering and comprehension stack, and that most sites have room to improve there.