A year after its first look at AI in content marketing, Ahrefs has found more marketers checking what AI writes for them, and a small but growing group who say they barely check at all.

The SEO software company surveyed 301 marketers in August and published the results on its blog on 30 September, with a press release following on 1 October. Of those respondents, 93% use AI to create or help create content, up from 87% in 2025. Among the 280 AI users, 7% said more than three-quarters of what they publish is “pure AI” content, which Ahrefs defines as material published without any editing or additional input. Last year the figure was 4%.

With adoption close to universal, the question for content teams is no longer whether to use AI but how much checking it gets before it goes out. That question now carries commercial weight. LinkedIn said in August that its detection classifiers had cut views of AI-generated posts by 40%, and Google’s search guidance says it judges content on quality and usefulness rather than on how it was produced.

What did Ahrefs’ 2026 survey find about unedited AI content?

Ahrefs’ 2026 survey found the group publishing mostly unedited AI content rising from 4% to 7%, but from a low base. The 7% who say most of their output is pure AI remain a small minority; on Ahrefs’ base of 280 AI users, that equates to roughly 20 people. The more telling shift sits at the other end of the scale. The share of AI users who publish no pure AI content at all fell from 50% to 40%.

The rest of the data points the same way. Some 69% of all respondents said they now publish a larger share of AI content than a year ago, rising to 74% among AI users. Speed is the main reason given: 85% of AI users said they produce content faster, with 47.9% describing it as much quicker and 37.1% as slightly quicker. Only 2.5% said AI had slowed them down.

The tasks handed to AI have widened too. Use for data analysis rose from 28% to 60%, image creation from 19% to 43% and keyword clustering from 23% to 45%. Claude was the most used tool among AI users at 83.2%, ahead of ChatGPT at 75.7% and Gemini at 57.9%, though Ahrefs cautions that it changed the question format this year, so tool figures are not a like-for-like comparison with 2025. A further 68% of AI users said they work with skill files, the reusable instruction sets that tell a model how to handle a recurring job, and 55% build their own, a pattern that echoes the homebrew AI tooling now absorbing hours inside marketing teams.

Two caveats limit how far the headline number can be pushed. Ahrefs does not say whether “more than three-quarters” is measured in words, articles or pages. And the report notes that last year’s survey had more respondents, advising that year-on-year changes be read as high-level signals rather than precise movements.

How many marketers check AI-generated content before publishing?

Ahrefs’ press release puts manual checking of AI-generated content at 89%, up from 80% in 2025. The figure appears in the release and in eMarketer’s coverage, but not on the published report page, which says only that manual self-checking is the most common method and that fewer respondents than last year skip checks altogether.

Trust figures on the report page tell a consistent story. Two-thirds (66%) of respondents said they trust AI-generated content only after a human has reviewed it. Just 2.7% said they trust it with no review. The release adds that 43% work with subject-matter experts on AI content and 33% use an editor.

Ryan Law, Ahrefs’ director of content marketing, said in the release that growing confidence in AI does not mean marketers are “handing over the keys”. He told eMarketer that growing executive buy-in and confidence in AI content are what drive the greater need for scrutiny and oversight.

The newest method draws doubts even inside Ahrefs. The share of respondents using AI to check AI-generated content doubled, from 14% to 29%. Louise Linehan, a content marketer at Ahrefs, described her own doubts about the approach in comments reported by PPC Land: an AI quality check that flagged problems left her wondering “what else AI had got wrong or missed”.

What does the Ahrefs survey not show?

The Ahrefs survey does not measure outcomes. It records what marketers say about their own practices, not audited content, and it contains no data on whether edited or unedited AI content performs differently in rankings, traffic or conversions. It cannot support a claim that either approach wins in search.

Google’s position offers no shortcut either. Its Search Central guidance says appropriate use of AI is not against its guidelines, while using automation mainly to manipulate rankings breaches its spam policies. The guidance advises creators to check AI output for accuracy and quality, which places the burden on the review step rather than on the production method.

Marketers’ own view of quality is shifting but still favours people. On the report page, 56% of AI users said human-written content is better than their AI content. The release reports that 24% now rate AI content as higher quality than human writing, up from 14% in 2025.

Disclosure has barely moved. Only 22% of AI-using respondents tell website visitors that AI helped create their content, up from 16% last year, which leaves 78% saying nothing.

Ahrefs has not published the respondents’ geography, recruitment method or industry split, and has not said how the 7% figure is measured.