How answer engines read a page.

An unlocked page is treated as untrustworthy and often skipped.

Serve HTTPS and redirect http to https.

The first string models quote when they name the source.

15–70 characters. Name the person or firm and the specialty.

Overviews pull this sentence when they summarize the page.

50–160 characters. One or two sentences.

Models treat the H1 as the page claim. Five H1s split the claim.

Exactly one H1. Put the practice statement there.

H2s are the outline answer engines walk.

At least three H2s that name real topics.

Question headings extract; slogans do not.

Make at least two H2s actual questions.

A self-contained 40–80 word answer is a strong citation predictor.

Put a direct answer in the first paragraph under the H1.

JSON-LD that parses and matches the page type beats unused types.

Person or Organization plus service type; articles need dates.

Hidden FAQ JSON-LD without on-page questions is discounted.

Publish 3–5 real questions and mark them as FAQPage.

og:title, og:description, and og:image are what cards cite.

Set all three on every public page.

No viewport meta and the page is treated as desktop-only.

Add width=device-width viewport.

Bullets extract as steps more often than paragraphs.

Use real ul/ol for process and proof.

Name, license, phone, and address decide source vs brochure.

Show a named person, license or firm, phone or email, and street address.

LinkedIn, YouTube, and profiles tie one person to one entity.

Add sameAs in Person schema and visible profile links.

Nav chrome is not the same as links into guides.

Link into topic pages, not only listing search.

alt="image1" does not attribute a photograph.

Specific alt on content images.

No date looks like a 2019 template.

datePublished and dateModified; visible last-updated on guides.

www vs bare vs IDX hosts split the entity.

One canonical that matches the public hostname.

Pages that cite .gov and primary sources get cited.

Link out to court, code, or regulator pages.

Blocking GPTBot or default-deny robots.txt blocks citation.

Allow Googlebot and live-answer bots.