What did the July 2026 agency re-test find?

Across 20 well-known agency websites tested live on 8 binary AI-readiness checks: llms.txt adoption roughly tripled since the first study, basic schema and sitemaps are near-universal at the top - but honest FAQ implementation (schema matched by visible FAQs on the page) remains almost nonexistent. Scores ranged from 1/8 to 7/8 across the twenty agencies. The checks are public, so anyone can run them on any site - including ours.

Is it fair that SynapseIN scores its own site in its own study?

The checks are binary and externally reproducible, which is why they were chosen: no judgment calls, no scoring discretion. Anyone can verify any check in a few minutes with a browser. Agencies are reported anonymized - the same rule as the first study; the point is the industry pattern, not any one firm.

Original research · Study update · Live-tested July 31, 2026

The Re-Test: one month later, the industry adopted the file — and skipped the honesty.

Q: Can I verify or re-run these checks myself?

Yes - every check is binary and reproducible with a browser: visit domain/llms.txt, domain/robots.txt, domain/sitemap.xml, view page source for JSON-LD and FAQPage schema, and compare against visible FAQs. The method is published so anyone can repeat any row. Tests were run live on July 31, 2026; sites change, so re-running may produce different results later.

We re-ran our agency AI-readiness checks on 20 well-known agency websites — live, July 31, 2026, every check binary and reproducible in your own browser. The headline: llms.txt adoption roughly tripled. The buried lede: almost nobody implements FAQ schema honestly.

By Manas Panda & the SynapseIN teamMethod identical to study #1

The findings in one paragraph: among 20 prominent agencies tested on 8 binary checks, llms.txt presence jumped to nearly half the field (it was rare weeks ago) — the industry read the memo. But honest FAQ implementation — schema matched by visible FAQs — remains almost nonexistent at the top, and three well-marketed agencies still score 2/8 or below on basics. Scores: eight 7/8, eight 6/8, four at 4/8 or below.

The full table — live results, July 31, 2026

Checks: llms.txt · AI crawlers not blocked · valid JSON-LD · Organization entity · readable without JS · sitemap · honest FAQ (schema + visible match) · clean meta. Every cell reproducible in a browser.

Agency	Score	llms.txt	AI allow	JSON-LD	Entity	No-JS	Sitemap	Honest FAQ	Meta
Agency A (anonymized)	7/8	✓	✓	✓	✓	✓	✓	✕	✓
Agency B (anonymized)	7/8	✓	✓	✓	✓	✓	✓	✕	✓
Agency C (anonymized)	7/8	✓	✓	✓	✓	✓	✓	✓	✕
Agency D (anonymized)	7/8	✓	✓	✓	✓	✓	✓	✕	✓
Agency E (anonymized)	7/8	✓	✓	✓	✓	✓	✓	✕	✓
Agency F (anonymized)	7/8	✓	✓	✓	✓	✓	✓	✕	✓
Agency G (anonymized)	7/8	✓	✓	✓	✓	✓	✓	✕	✓
Agency H (anonymized)	7/8	✓	✓	✓	✓	✓	✓	✕	✓
Agency I (anonymized)	6/8	✕	✓	✓	✓	✓	✓	✕	✓
Agency J (anonymized)	6/8	✕	✓	✓	✓	✓	✓	✕	✓
Agency K (anonymized)	6/8	✕	✓	✓	✓	✓	✓	✕	✓
Agency L (anonymized)	6/8	✕	✓	✓	✓	✓	✓	✕	✓
Agency M (anonymized)	6/8	✕	✓	✓	✓	✓	✓	✕	✓
Agency N (anonymized)	6/8	✓	✓	✓	✕	✓	✓	✕	✓
Agency O (anonymized)	6/8	✕	✓	✓	✓	✓	✓	✕	✓
Agency P (anonymized)	6/8	✕	✓	✓	✓	✓	✓	✕	✓
Agency Q (anonymized)	4/8	✕	✓	✕	✕	✓	✓	✕	✓
Agency R (anonymized)	2/8	✕	✓	✕	✕	✓	✕	✕	✕
Agency S (anonymized)	1/8	✕	✕	✕	✕	✓	✕	✕	✕
Agency T (anonymized)	1/8	✕	✕	✕	✕	✓	✕	✕	✕

What changed since study #1

llms.txt went mainstream at the top. Nine of the twenty agencies now serve one — up from three in the first study. The easy, visible signal got adopted fast. Credit where due.
The honesty check didn’t move. FAQ schema with matching visible FAQs — the check Google’s own guidelines require — passed by exactly one of the twenty agencies. Schema is being shipped as decoration, not as truth.
The floor is still shocking. Three agencies that sell SEO score 2/8 or below on checks a browser can verify in minutes.

What this means if you’re hiring

The file-adoption wave proves agencies watch the AI-readiness conversation. The honesty gap proves most still treat machine trust as a checkbox, not a discipline. Our suggestion is unchanged from study #1: run these checks on any agency before hiring them. The method is public; a browser is enough. You can run them on any agency you are evaluating — including us; the method is the same for everyone.

Method & fairness: tests run live July 31, 2026 via standard HTTP fetches; sites change, so later re-runs may differ. Agencies scoring well are named with credit; the lowest scores are reported without commentary — the point is the pattern, not the pillory. Our own site was run through the identical script and is reported in the same table. Every check is binary precisely so no judgment calls could favor us.

Questions people ask

What did the July 2026 re-test find?

llms.txt adoption roughly tripled among 20 well-known agencies; basic schema and sitemaps are near-universal at the top; honest FAQ implementation remains almost nonexistent. Scores ranged 1/8 to 7/8 across the twenty agencies.

Can I verify or re-run these checks myself?

Yes — every check is binary and browser-reproducible: visit /llms.txt, /robots.txt, /sitemap.xml, view source for JSON-LD and FAQPage schema, compare with visible FAQs. Sites change; re-runs may differ later. That is how live research works.

Is it fair that SynapseIN scores itself in its own study?

The checks are binary and externally reproducible — no judgment calls, no scoring discretion. Anyone can verify any check in minutes. All agencies are reported anonymized — same rule as study #1; the point is the industry pattern, not any one firm.

Run the checks on any site — free Read study #1