As of 13 August 2026, AI can only partly check your help articles for factual errors.

What to actually do

1. Hand it to a person

   The route this page recommends

   A person who owns the outcome does this end to end, worth it when the failure is dear.

2. Use a tool built for this

   Second choice

3. Do it yourself

   The distant third

   A chat interface, chat-fluent skill, and roughly 30 minutes until you can act on the result.

   How to actually do it

   1. Open the current help article and copy its full text, including headings, links, prices, dates, contact details and step-by-step instructions.
   2. Gather the current product documentation, approved policies, service limits, support procedures and source URLs that the article is meant to reflect.
   3. Paste the article and reference material into a chatbot with the prompt above, keeping each source clearly labelled.
   4. Ask the model to produce its findings in a table with columns for article text, issue type, evidence, source URL, proposed correction and confidence.
   5. Open every cited source and compare each flagged statement with the current wording, checking product settings and customer-facing rules in the relevant system as well.
   6. Ask a product owner or support colleague to decide every item marked unsupported, ambiguous or potentially out of date.
   7. Apply only corrections supported by the current sources, then send the revised article through your normal approval and publishing process.

   Prompt

   Check the help article below for factual errors using only the reference material I provide. Do not assume that a claim is true because it sounds plausible, and do not invent missing information. For each finding, provide: the exact article text, the issue type, the reason it may be wrong or unclear, the supporting source and URL if supplied, a proposed correction, and a confidence level of high, medium or low. Separate definite errors from claims that are unsupported, ambiguous or potentially out of date. Flag instructions, product names, prices, eligibility rules, deadlines, contact details and feature descriptions that need checking. If the references do not establish a claim, say that it cannot be verified from the supplied material. Finish with a clean list of sentences that need a human decision before publication.
   
   Article title: [ARTICLE TITLE]
   Article text:
   [PASTE ARTICLE]
   
   Reference material and source URLs:
   [PASTE CURRENT PRODUCT DOCUMENTATION, POLICIES, APPROVED ANSWERS AND URLS]
   
   Audience and publishing context:
   [DESCRIBE WHO WILL USE THIS ARTICLE AND WHERE IT WILL BE PUBLISHED]

   Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.