01
Clarity
One clear topic
A descriptive title and H1 state the topic and purpose. Pages that try to rank for five intents end up being the best answer for none of them, and AI systems pull passages from pages whose purpose is obvious.
Do this: Write the title as the question or task the reader brought with them. If you can't summarize the page in one sentence, split it.
02
Clarity
Answer up front
Answer the main question early, with enough context to stay accurate. The opening paragraph is the passage most likely to be lifted into an AI answer, so it has to stand on its own without the rest of the page.
Do this: Put a 2 to 3 sentence answer directly under the H1. Include the qualifier that keeps it true (who it applies to, when, under what conditions).
03
Evidence
Credible identity
Show the publisher, relevant expertise, and meaningful dates. Readers and models both weigh who is talking. An anonymous page with no update date gives them nothing to weigh.
Do this: Add a real byline linked to an author page, a short line on why that person is qualified, and an updated date that only changes when the content does.
04
Clarity
Headings that guide
Follow the reader's questions, tasks, and next steps. Headings act as a table of contents for humans and as section labels for machines that chunk your page into passages.
Do this: Read only your H2s in order. They should tell the full story of the page. Swap clever headings for literal ones.
05
Evidence
Original value
Add firsthand evidence, demonstrations, research, or useful analysis. If your page says what fifty other pages say, a model has no reason to cite yours. Original data, screenshots, and tested results are what make a source worth naming.
Do this: Add one thing nobody else has: your own test, a real number from your work, a photo you took, or a worked example.
06
Evidence
Facts with context
State names, units, dates, conditions, and definitions clearly. "Fast" means nothing. "Loads in 1.8 seconds on a mid-range phone over 4G" can be checked, compared, and quoted.
Do this: Hunt for vague adjectives and replace each one with a number, a date, or a named thing.
07
Evidence
Evidence beside claims
Place credible sources near claims. Explain method and limitations. A sources list buried at the bottom helps a little. A source sitting right next to the sentence it supports helps a lot more, for readers and for any system checking the claim.
Do this: Link primary sources inline. When you ran your own test, add one sentence on how you did it.
08
Experience
The right format
Use tables, steps, and lists when they improve understanding. Format isn't decoration. Comparisons belong in tables, processes belong in numbered steps, and concepts need a plain explanation with an example.
Do this: Use the format matcher below to check each section against the task it serves.
09
Experience
Visuals that explain
Add useful captions and accessible descriptions. Keep key facts as text. A chart can make a point land, but if the only copy of a number is baked into an image, most crawlers and models never see it.
Do this: Write alt text that says what the visual shows, and repeat its key takeaway in the body copy or caption.
10
Evidence
Tradeoffs and limits
Explain assumptions, uncertainty, exceptions, and who the advice fits. Pages that admit limits read as more trustworthy, and they protect you from being quoted out of context in situations your advice was never meant for.
Do this: Add a short "When this doesn't apply" or "Who this is for" block.
11
Access
Helpful internal links
Use descriptive links and make this page reachable from related pages. Internal links are how crawlers discover pages and how they understand which page owns which topic. An orphan page is close to invisible.
Do this: Link to this page from at least three related pages, using anchor text that describes what's on the other side.
12
Experience
Useful follow-up answers
Address real next questions and offer a useful next step. Does every page need an FAQ? No. Only when it answers genuine reader questions. Padding a page with invented questions adds noise, not authority.
Do this: Pull follow-up questions from Search Console queries, sales calls, and support tickets. Answer only the ones people really ask.
13
Access
Agent-ready controls
Use real buttons, labeled fields, stable layouts, and clear feedback. AI agents are starting to fill forms and complete tasks for people. The same accessible markup that helps a screen reader helps an agent understand what a control does.
Do this: Use native <button> and <label> elements, avoid layout shifts, and show validation and completion messages as text.
