Audit your top 20 help articles in an afternoon

Time to set up

One afternoon

Cost

£0

Driver

Technology

The problem

Your AI answers from your help centre. Increasingly, so do the tools customers use on their own, like ChatGPT. Most help centres have a long tail nobody reads and a short list everybody does. If one of the busiest articles is wrong, every customer who asks gets the wrong answer, at any hour, until someone notices.

Gartner found the most common reason self-service fails, in 43% of cases, is that customers can’t find content relevant to their issue.

The fix

1. Pull the usage list. Export the articles your AI used or customers viewed most in the last 30 days, and take the top 20.
2. Add two columns. The date each article was last updated, and who owns it. Leave the owner blank where there isn’t one.
3. Check each one against the live product. Open the product beside the article and follow the steps. Mark it correct, needs a fix, or retire.
4. Fix or retire on the day. Make small corrections straight away. Anything bigger becomes a ticket with a date. Unpublish anything you retire, so the AI can’t find it.
5. Name an owner and a review date for every article that stays.
6. Repeat monthly for the top 20. The list will change as your product does.

Tools to try

• Your knowledge base analytics (Zendesk, Intercom, Freshdesk) for article views and last-updated dates.
• Your AI agent’s reporting on which articles it used to answer, if it has it. That’s the most accurate usage list you can get.
• A shared spreadsheet for the audit itself. The habit matters more than the tool.

How to measure it

Track the share of your top 20 articles reviewed in the last 30 days (aim for all of them), and escalations where the AI couldn’t answer on the topics those articles cover, before and after the audit.

Why this matters: Nobody owns your help centre, and your AI shows it