Best practice · Checklist
AI visibility best practices: the 12-point checklist
By Kristian Stig Henriksen · Dear Future
Best practice for AI visibility is not a secret technique. It is twelve things that make a page easy for an AI assistant to find, understand and cite, and a measurement that shows whether it works. The list here is the one we work by ourselves. Every point is one thing you can tick off, with a link to the guide that goes deeper.
Your own pages
- 01
One short answer at the top
The first two sentences on the page answer the page's question, so they can be quoted alone, without the rest of the page.
Go deeper: Get recommended by ChatGPT
- 02
One definition per concept
Every page defines the concept it is about in one sentence and uses the same wording everywhere on the site.
Go deeper: What is GEO
- 03
Questions as headings, and an FAQ in the right words
The headings are the questions people ask, in their own words. The FAQ block is mirrored verbatim in structured data, never invented for the occasion.
Go deeper: AI SEO in Denmark
- 04
Date and author
Every page shows when it was last updated and who wrote it, and the structured data says the same.
Go deeper: The method
Sources and mentions
- 05
Consistency across sources
The same description of what you do, for whom and where, on your site, LinkedIn, directories and at partners. Models weight agreement between sources.
Go deeper: Influence what AI says
- 06
Mentions at the sources the models already cite
Find them in your category: industry overviews, trade media, partners. Get named there. It is not link building; it is being present where the answer is fetched from.
Go deeper: Share of voice in AI
- 07
Your own sources on the page
Claims with numbers carry a source with a link. That is how the page itself becomes a source a model dares to build on.
Go deeper: Confidence interval
Technology
- 08
Can be fetched
The page responds with its full text without requiring JavaScript, a login or consent first, and robots.txt does not shut AI crawlers out.
Go deeper: AI Overviews and AI Mode
- 09
A sitemap with dates and llms.txt
The sitemap says when each page last changed, because that is what Google schedules from. An llms.txt lists the pages with one line each for the crawlers that read it.
Go deeper: LLM SEO
- 10
Structured data that matches what is visible
Article, FAQPage, Organization and breadcrumbs, and never anything in schema that is not on the page.
Go deeper: What is MCP
Measurement
- 11
A baseline and a control group
Measure before you change anything, and leave some questions untouched, so a change can be attributed to the work and not to chance.
Go deeper: Check your AI visibility yourself
- 12
Read in windows, with an interval
One day is noise. Compare weeks, and show the uncertainty on the number, so a movement can be told from a coincidence.
Go deeper: How to measure AI visibility
What is not on the list
Tricks that try to write the brand into the answer around the sources: hidden text, invented reviews, pages that ask about themselves. They do not work in answers with search, and they pollute the measurement. Best practice is the same as good SEO plus measurement; the difference is that the yardstick is the answer, not the list.
From checklist to numbers
Want to know how many of the points already work for you?
Get an AI visibility analysis: we measure daily who the assistants name and cite in your category, and what separates you from those who are named.
Get the analysis →Frequently asked questions
Where do you find best-practice recommendations for AI visibility?
Google's own documentation on AI features says there is no special optimisation and that the same pages that rank become sources. The rest is experience: the twelve points on this page collect what makes a page easy to find, understand and cite, and each of them links to a guide that goes deeper.
Which methods are recognised for improving a brand's visibility in AI answers?
Clear pages with a short answer at the top and one definition per concept, questions as headings with an FAQ in people's own words, a date and an author, consistent descriptions across sources, mentions at the sources the models cite, technical accessibility with a sitemap and llms.txt, and measurement with a baseline, a control group and an interval.
What are the most important guidelines for optimal AI visibility?
Answer customers' questions rather than talking about yourself, make the pages easy to fetch and understand, be consistent across sources, and measure the result over weeks rather than days. Avoid tricks that write the brand into the answer around the sources; they do not work in answers with search.