The Five-Layer AEO Playbook for B2B Service Firms

Score each layer against its pass threshold. Twenty checks, one afternoon. Nothing you type leaves this page, nothing is stored and nothing is sent anywhere. Print or save to PDF to keep your result.

1 Entity sentence

Pass threshold: one sentence, word for word identical in all three places.

A single sentence names what the firm does and who it does it for, in under 25 words.
That exact wording appears on the homepage.
That exact wording appears on the about page.
That exact wording appears in the LinkedIn company description.

Fail if any of the three paraphrases it. Paraphrase is what stops a model quoting you.

Your entity sentence

2 Category claim

Pass threshold: a real label a buyer would type, used in the H1, claimed no more than twice.

The category label returns real results when searched, so it is not invented.
The label appears in the H1 of the relevant service page.
The firm claims two categories at most across the whole site.
The same label is used in the entity sentence from layer 1.

A coined category with no search behind it gives a model nothing to match a buyer question against.

Your category label

3 Off-site corroboration

Pass threshold: five sources on five domains you do not control, each carrying the firm name next to the category label.

Source 1 is live and names the firm alongside the category label.
Source 2 is on a different domain from source 1.
Source 3 is on a third domain.
Source 4 is on a fourth domain.
Source 5 is on a fifth domain.
None of the five is a guest post the firm wrote about itself.

Directory listings, association member pages, podcast episode notes, conference speaker bios and client-side press all count. Low traffic does not disqualify a source. Review platforms lost most of their audience and kept their citation weight.

Your five domains

4 Evidence pages

Pass threshold: three case studies, each with a client type, a starting number and an ending number in the body text.

Three case studies exist as crawlable pages, not as gated PDFs.
Each states the client type or sector in words a buyer would use.
Each states a starting number and an ending number as text, not inside an image.

Numbers rendered only inside a graphic are invisible to a text fetcher, so the one quotable fact on the page never gets quoted.

Your three case study URLs

5 Machine access

Pass threshold: fetchers are unblocked and the main content renders without JavaScript.

robots.txt does not block GPTBot or OAI-SearchBot.
robots.txt does not block PerplexityBot, and the Google-Extended decision was made deliberately.
Service pages render their main body text with JavaScript disabled.
Server logs have been checked for those user agents rather than assumed.

Google-Extended is a robots.txt token, not a user agent. It will never appear in a log, so its absence there proves nothing.

Fetchers seen in your logs

0 / 21

Tick the checks you can evidence today.

Score	What it means	Do this next
0 to 7	Nothing for a model to corroborate. Expect not to be named.	Layers 1 and 2 only. Do not produce content yet.
8 to 14	The claim is consistent but stands alone on your own domain.	Layer 3. This is where the calendar time goes.
15 to 18	Corroborated. Naming becomes plausible and worth measuring.	Baseline your share of model before changing more.
19 to 21	Every layer carries. Remaining variance is engine volatility.	Re-measure monthly across several assistants.

Companion template to AEO for B2B Service Businesses: A Practical Playbook. The score bands are peppereffect's working thresholds, assembled from client audits and published correlational research. They are a starting discipline, not a tested dose-response curve. Version 16 September 2026.