Here is the deal on AI content: Google has said, on the record, that it is fine, as long as it is helpful and made for people. The March 2024 update then wiped out a pile of sites that used AI to publish at scale with nothing new to say. Both things are true.
What Google's policy actually says
Google rewards high-quality content "however it is produced". It targets "scaled content abuse": producing many pages primarily to game rankings, with little value, whether by AI, humans, or both. The tool is not the problem. Publishing filler at volume is.
Where the line sits
| Usually fine | Usually filtered |
|---|---|
| AI drafts, then a human adds experience, data and edits | Publish the raw AI output, at volume, unreviewed |
| AI for outlines, research summaries, first passes | Hundreds of near-identical pages from one template prompt |
| AI to scale a proven, useful format | AI to chase every keyword with no first-hand input |
| Named author who stands behind the page | No author, no accountability, no originality |
How to use AI without getting hit
- Start from your own experience, data or opinion. Use AI to structure and speed it up, not to source it
- Add something no other page has: a test result, a screenshot, a number, a take
- Edit hard. Cut the throat-clearing and the generic advice
- Put a real author on it, with a bio and a track record
- Publish at a pace you can keep quality at, not the pace the tool allows
Bottom line: AI is a fine writing assistant and a terrible strategy. The sites that got hurt did not lose because they used AI. They lost because the pages had nothing in them.
What the March 2024 update actually did
Tanbir Habib Riyad · SEO Trainer & Content Strategist
We use AI for the first 60% of a draft: structure, research summaries, filling obvious gaps. The last 40% is a human adding the test result, the screenshot, the opinion. That last 40% is the whole reason the page ranks.
Short case study
An affiliate site hit in March 2024
The site had published 400 AI-written "best X" posts in a year with no first-hand testing. Traffic dropped roughly 80% overnight. We cut the library to the 70 pages with real demand, added hands-on testing, photos and a named reviewer to each, and consolidated the rest with redirects.
Traffic returned to about 55% of its former peak over the next two core updates, on one sixth of the page count and with far better conversion.