Ahrefs' AI Content Workflow Keeps the Editor in Charge

Ahrefs marketer Si Quan Ong’s October 16 guide shows content teams how recurring updates can stop short of automatic publication. An editor keeps the final decision. A small team can use the example to separate repeatable maintenance from decisions about what readers should see.

This is a more detailed guide to existing work. Ong first described the Data Refresh Hub on June 25, when it had already been running for two months. The new account isn’t a product launch or evidence that a marketing department suddenly became autonomous.

Ong, an SEO & Marketing Educator at Ahrefs, sets a firm rule for publication.

“I never let AI publish automatically for me.”

The current guide describes monthly updates across 12 datasets, including new tables and charts, delivered as WordPress drafts for review. The June account listed 14 datasets; neither explains the difference. Treat those figures as descriptions of each account, rather than proof of expansion or contraction.

Our earlier reporting on Anthropic’s marketing operations examined a similar review checkpoint. Ong’s example adds a useful question for content owners: what did the system leave out?

Inspect what disappeared

Ong advises the builder to:

“keep what it rejected”

A polished table can conceal a bad exclusion. When a system filters a list, checking a correct entry won’t reveal a valid result it removed. An editor needs access to those removals to judge whether the cleaning rules match the article’s purpose.

That changes what a finished update should contain. Alongside the proposed page, a reviewer should be able to inspect the underlying input and the exclusion decisions. Otherwise, the person approving the article is working from the same incomplete view as the reader.

For a lean team, this is a manageable first project: choose an existing update with clear source data and a known destination. Define the decisions the editor makes each month before asking software to make them repeatedly. A proposed automation that can’t explain an exclusion needs further testing before it earns publishing access.

Make the review a real checkpoint

The publishing software can support that division of work. WordPress’s own documentation distinguishes draft and pending posts from published ones in its update interface. A team connecting an AI workflow can deliberately send the result into review instead of requesting publication. The available setting alone doesn’t prove a particular integration has been configured safely.

Another vendor’s account illustrates why the review needs substance. In Anthropic’s July 8 description of its marketing operations, a separate proofreading step checks report numbers against verified sources. Campaign operator Annabel Custer reviews results before they ship. That provides a concrete comparison for the design of the checkpoint, rather than independent confirmation of Ahrefs’ results.

Our coverage of marketers’ concerns about AI and brand voice adds another review question. Does the revised text still sound like the company? A review button is useful only when its owner has something meaningful to check. Give that person the source data, explain the requested change, and preserve a way to compare it with the previous version.

Budget for the work around the workflow

Ahrefs sells the products involved, so the account also serves a commercial purpose. The June post reported a time saving, but it wasn’t an independent time study. Neither account establishes a measured return for another content team. The fresh guide also acknowledges that the automation still needs maintenance.

Start with one update that someone already owns. Keep a record of corrections and the time spent reviewing each run. Evaluate the total effort, including repairs, before extending the same setup across the content operation. The editor approving publication should be able to explain which material the system rejected and why it belongs outside the finished page.