We checked 18 B2B content marketing blogs for AI-answer readiness. Then we checked our own.

GEO (Generative Engine Optimization) is the practice of making content easy for AI answer engines — ChatGPT, Perplexity, Claude, Google AI Overviews — to find, parse, and cite. A lot is being said about it and not much is being measured. We ran a small, mechanical check against 20 well-known SEO and content-marketing sites — the people who write the guides everyone else follows — using WellCited, the free WordPress plugin behind this site. We included our own site in the results.

This is not a ranking and not an accusation. It is six yes/no or three-state checks against one article per site, on one day. Read the methodology before you read the numbers.

Methodology

• 20 target domains, chosen for name recognition in SEO/content marketing (Ahrefs, Semrush, Moz, HubSpot, Backlinko, Search Engine Journal, Search Engine Land, Yoast, and others), plus wellcited.app itself.
• One article per site — usually the site's own flagship "what is SEO" / core-topic guide, fetched live. This is a snapshot of one page, not a whole-site score.
• 18 of 20 pages were reachable at measurement time; 2 (growandconvert.com, clearscope.io) failed to resolve from our test environment and are excluded — that's a limitation of our setup, not a claim about those sites.
• One data point per site. A different article on the same domain could score differently. Treat this as a snapshot, not a verdict.

What we checked

All mechanical, all reproducible with the same tool:

• llms.txt — does the site publish a plain-text content index at /llms.txt?
• Question headings — is there at least one H2/H3 heading ending in "?"
• Short answer under heading — does a question heading have a 1–300 character paragraph immediately under it (the kind of extractable, quotable answer an AI engine can lift directly)? A long lead-in paragraph before the real answer does not count.
• Heading hierarchy — do H2–H6 headings nest without skipping a level?
• Author / date signals (aggregate only, not published per-site below — more implementation-sensitive and less black-and-white than the others) — is authorship/date visible in the HTML itself, only in JSON-LD, or not present at all?

Results

Aggregate

Across the 18 measured pages:

• 9/18 serve an /llms.txt file.
• 11/18 have at least one question-style heading.
• 9/18 pair a question heading with an immediate, short (≤300 character), extractable answer.
• 10/18 have a valid, non-skipping heading hierarchy — the rest either skip a level or start below H2.
• Author signal: 5/18 visible in HTML, 12/18 only in JSON-LD (invisible to a human reader), 1/18 neither.
• Date signal: 4/18 visible in HTML, 12/18 JSON-LD only, 2/18 neither.

The pattern that stands out most: structured data (JSON-LD) is common, but it's mostly doing the author/date work alone — 12 of 18 sites mark up who wrote a piece and when only in a hidden schema block, not in text a person would actually see on the page.

Our own result

At the time of the original audit (2026-08-14), wellcited.app didn't come out ahead of the sites we measured either. Our FAQ content was marked up correctly in JSON-LD (FAQPage, Question, Answer), but it wasn't built from real <h2>/<h3> question headings in the page HTML, so it read as question headings: no and heading hierarchy: invalid under the same mechanical check we ran on everyone else.

We've since fixed it: the FAQ section now uses real <h3> question headings with a short answer paragraph directly underneath each one, and the heading hierarchy on the page no longer skips a level. The table and aggregate numbers above reflect that update. We're leaving this paragraph in rather than quietly editing the result, because the whole point of the study is that these are checkable, fixable things — not a permanent verdict.

What this does and doesn't show

It doesn't show who ranks best in ChatGPT or who gets cited most — that requires tracking actual AI answers over time, which is a different (and much harder) measurement. It does show that even among sites whose entire business is content and search visibility, machine-extractable structure — short direct answers, visible authorship, consistent heading nesting — is inconsistent. Half the sites here fail a heading-hierarchy check that costs nothing to fix.

WellCited runs this exact check (plus 19 on-site audit rules) for free, entirely inside your own WordPress admin — no API key, no external request, no account. If you want to see where your own content stands, get the free plugin.

Methodology note: measured with WellCited Pro's Competitor Check tool against one article per domain on 2026-08-14.

Published: 17 August 2026