Write articles the AI retrieves well
To get an article retrieved by the AI, do four things: write in the customer’s own words, put the answer in the first paragraph, keep the article to one topic, and use the keywords field for synonyms, product terms, and error codes. These habits target both retrieval paths — keyword and vector — at once, and they help every AI surface plus the public help center search, because all of them use the same engine.
Write for both paths at once
There is no retrieval mode to choose when authoring. Every article is searched by two parallel paths — a keyword path that matches literal words and a vector path that matches semantic meaning — so write for both simultaneously: clear prose for semantic matching, and the customer’s exact words for keyword matching.
Write the customer’s words
Embeddings catch paraphrases, but a chunk that uses the customer’s exact phrasing wins on both paths simultaneously. Phrase headings and key sentences the way a customer would ask, not the way the product team describes the feature internally.
Lead with the answer
The first chunk of an article is often the most-retrieved chunk. Don’t bury the lede — state the answer or outcome in the opening paragraph, before any background or setup.
One topic per article
Articles are split into chunks of a few hundred words, and results are deduplicated so each article appears at most once. Two topics in one article means one chunk per topic, but the article ranks at most once — the topic with weaker chunks loses. Split multi-topic articles into focused ones.
Use the keywords field
The keywords field is included in chunk embeddings, so it helps semantic and vector matching without forcing awkward synonyms into the prose. Use it for synonyms, product terms, error codes, and common misspellings you want the AI to understand without polluting the article body. Title and summary get weighted higher in the keyword path, and the body still grounds the answer, so make sure the customer’s core terms appear there too.
Keep articles current
Stale articles don’t carry an explicit penalty in retrieval, but outdated content erodes the AI’s accuracy. The quarterly review pass is the cheapest fix.