Get started
Generative Engine OptimizationContent

The GEO Content Checklist: 12 Checks Before You Publish

A concrete pre-publish checklist for content you want generative engines to quote. No theory, twelve checks you can run in ten minutes.

Visibilitas Team 3 min read
The GEO Content Checklist: 12 Checks Before You Publish

Most GEO advice is strategic and therefore hard to act on. This is the opposite: a list you can run against a draft before it goes out. It takes about ten minutes and catches the problems that keep otherwise good pages out of answers.

Structure

  1. Every heading is a question or a claim, not a label. "Pricing" is a label. "What does it cost for a five-person team?" is a question somebody types. Headings are how a retriever decides which chunk answers a prompt.
  2. The paragraph under each heading answers it completely, alone. Read it with the rest of the page covered up. If it depends on the section before it, a model that lifts it will lift something incoherent, so it will lift someone else's instead.
  3. The first paragraph of the page answers the title. Inverted pyramid. Context and nuance come after, not before.
  4. No pronoun opens a section. "It also supports..." is unquotable out of context. Name the subject again. Slightly repetitive for a human reader, decisive for a machine one.

Substance

  1. At least one falsifiable fact per section. A number, a version, a limit, a date, a price. "Highly scalable" cannot be quoted. "Handles 40 projects per workspace on the Agency plan" can.
  2. Say who it is not for. Balanced sources get cited on comparison prompts, which are the commercially valuable ones. A page that claims to suit everybody is evidence of nothing.
  3. Name the alternatives honestly. If your page is the only one in your category that never mentions a competitor, it reads as promotional and gets treated accordingly.
  4. Every claim has a visible source or is your own first-hand data. Attributable claims survive fact-checking passes. Unattributed assertions get dropped when a model finds them contradicted elsewhere.

Machine readability

  1. The content is in the served HTML. Fetch it with curl. If your text is not in the response body, it does not exist for most AI crawlers.
  2. The page has a visible date and a named author, and the same values appear in Article markup. See Structured Data for AI Answers.
  3. The AI crawlers are allowed. Check robots.txt for GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot and Google-Extended, and make each one a decision rather than an accident.
  4. Tables and lists carry the comparable facts. Specifications buried in prose are hard to extract reliably. A table of the same values is not.

The one that is not a check

None of the above tells you whether it worked. Publishing a page that satisfies twelve criteria and assuming the engines noticed is the same mistake as publishing for search and never looking at rankings.

Ask the engines your buyers' questions on a schedule and record what comes back: whether you were named, where in the answer, in what tone, and which URL was used. That last field is the useful one, because it tells you which page is doing the work so you can write more like it. The method is in How to Measure AI Search Visibility.

If you keep only one line from the list: write every section so it can be cut out of the page and still make sense. Nearly every other item on the checklist is downstream of that one.

Generative Engine OptimizationContent

← All posts

Keep reading