What 5,771 AI-assisted blog posts actually did in Google
The median article got 1 click in 90 days. The top 10% got 73% of all clicks. What the winners had in common, and what never won.
Everyone has an opinion about AI content. I wanted numbers.
So we followed 5,771 AI-assisted articles across dozens of brands in 11 industries, through 16 months of daily Search Console data. Each article was measured from its own publish date and ranked against the same brand's other articles, so a strong domain never flatters a format. The articles were drafted with AI, and some were revised after human feedback.
The power law
- The median article earned 1 click in its first 90 days. 49% earned none.
- The top 10% took 73% of all clicks. The top 1% took 22%.
If you judge a content program by its average article, it looks broken. If you judge it by its top 10%, it looks like a business. The job is finding and feeding that 10%.
What the winners had in common: a decision
Share of each format that became a top-20% article for its brand (20% would be chance):
- Rules and regulations explainers: 32%
- Best-of / ranked lists: 30%
- X vs Y comparisons: 27%
- A specific problem and its fix: 26%
- Cost and pricing breakdowns: 26%
- How-to guides: 17%
- Explainers: 17%
- Generic tips listicles: 15%
- Case studies: 9%
- "Why choose us" posts: 3%
Every winner helps someone choose: which one, which is better, is it allowed, how much, what's wrong with mine.
And 54% of what got published was explainers, how-tos and tips lists, which are the median formats.
Keyword difficulty mattered. Search volume didn't
- KD 0–9: 26% became top performers
- every band at KD 10 and above: 14–19%
- planned search volume: no relationship at all
Here's why volume doesn't matter: only 14% of articles ranked for their exact target keyword. 74% of clicks came from searches the article never targeted. You're ranking a topic. The focus keyword is a label.
The cluster is the real bet
For a typical brand, one topic cluster carried 37% of all clicks. Cluster choice explained about a quarter of the differences between articles for the same brand, more than any title or keyword trait.
So review clusters at 90 days. Put new articles into the one or two that are winning. Stop feeding the dead ones.
Core updates sorted decision pages from explainers
Same established articles, before vs after the May 2026 core update:
- buying guides: 0%
- cost pages: −13%
- rules explainers: −15%
- explainers and how-tos: −36 to −37%
The content an AI answer can replace lost the most. The December 2025 update went the other way: +69%.
AI tools are already searching your pages
Some queries in Search Console aren't typed by people. They're prompts:
context: location: … do not include location references in your response. question: what's the best way to compare lesson packages when schools price per class vs per term?
We found 119 across 84 articles. Average position 3.8. Clicks: zero.
They're the literal questions buyers ask AI, and the page the AI found. Filter your Search Console queries for context: or anything over 10 words and read them. Each one is a section you should write.
A note on impressions
Google confirmed that Search Console misreported impressions, CTR and average position from 13 May 2025 to 27 April 2026. Clicks were fine. If your report shows a big impression surge or collapse across that window, check clicks before you celebrate or panic.
What I'd do with this tomorrow
- Rebuild the calendar around decision content: comparisons, best-of lists, cost breakdowns, rules explainers, specific fixes.
- Choose KD under 10 and specific topics over volume.
- Give every close variant of the question its own section.
- Review clusters at day 90 and feed the winners.
- Read the AI prompts in your Search Console and answer them.
Full tables and method: the research page.