Workflow · AI-readiness and AEO
Content AI can find and quote
Real work Rebuilt from a Claude plugin I built for a content team
Answer engine optimisation (AEO) makes sure artificial intelligence (AI) assistants and AI search find your content, trust it, and quote it correctly. This workflow starts from the Content Design System, because most of its rules are the same signals AI uses. Then it adds the checks only AEO needs, and tests the result by asking Claude and Gemini real customer questions.
- Select Run the check.
- Read which articles fail and why, then pick a question to compare the answers.
- Approve the fixes, and the workflow asks the questions again.
Why the Content Design System comes first
Each rule on the left already does work for AI. The workflow checks these first, then adds four checks the system does not cover.
- Where an assistant will rank or quote you. Assistants change often, so the test runs again after every change and on a schedule.
- Whether an answer is right for every customer. The answer key comes from the product, and a person checks it.
- Anything about pages it cannot read. If the site blocks a crawler, the assistants never read the content at all.
Safeguards
What keeps the workflow safe to run on a real knowledge base.
- Checks are read onlyStages one to four read the articles and ask questions. Nothing changes until you approve the fixes.
- A fixed answer keyEvery answer is scored against a key written from the product, not against what sounds right.
- The same questions, before and afterThe test asks the same questions in the same way, so a better score comes from the content, not the question.
- Facts are checked against the productWhen two articles disagree, a person confirms the fact before either one changes.
- Real answers, named and datedEvery answer on this page is a real answer from Claude or Gemini, shown as given, with the date of the test.
- No made-up sourcesAn answer that cites an article that does not exist counts as wrong, even when the words sound right.
Take it to the next level
How to get results that match the best teams, and what to build next.
- Test more than one assistantRun the questions in Claude, Gemini, and the AI search tools your customers use, and compare how each one behaves.
- Use real questionsTake the questions from search logs and from what the chat bot handed to people, not from a list you wrote.
- Track it over timeRun the test every month and after every big content change, and show the trend on the audit report.
- Write for answersStart each article with a two-sentence answer an assistant can quote, then give the detail.
- Mark it upAdd structured data for questions and steps, and keep the last updated date visible on every page.
- Keep facts in one placeHold product facts such as deadlines and fees in one source, and pull them into every article, so they never disagree.
Rebuilt from a Claude plugin I built for a real content team. Keystone is an invented payments product, and every name, figure and conversation here is made up.