Ahrefs' second annual survey of 301 marketers finds 69% are publishing a larger share of AI-assisted content than a year ago — and among AI users specifically, that figure climbs to 74%. The survey, run by Glen Allsopp, Ahrefs' Head of Marketing Strategy & Research, also documents a sharp rise in adoption of specialized AI tasks and a persistent gap between how much AI content marketers publish and how much they disclose.

Adoption is now near-universal among respondents. Out of 301 responses, 280 marketers (93%) use AI to create or help create content, up from 87% in last year's survey. Among the roughly 7% who don't use AI, almost all cited accuracy as a concern, and more than half said they generally distrust AI-generated output.

The survey sits against a backdrop of accelerating mainstream AI usage. OpenAI reported 800 million weekly active users in October 2025, 900 million in February 2026, and passed 1 billion by the end of July.

Speed, not just volume

Last year's survey found AI users published 42% more content than non-users. This year, production speed is the standout metric: 85% of marketers using AI say they produce content faster since adopting it. 47.9% say they're much quicker, 37.1% slightly quicker, and just 2.5% say AI has slowed them down.

The biggest year-over-year shifts came in specialized tasks rather than drafting. Data analysis rose from 28% to 60%, image creation from 19% to 43%, and keyword clustering from 23% to 45%. Notably, these were among the least popular AI use cases in last year's data. Allsopp attributes the growth to a mix of improved models, shared workflows, and AI use becoming habitual.

Claude leads the tool list

Among AI users, 83% use Claude, 76% use ChatGPT, and 58% use Gemini. They're followed by Claude Cowork (42%), Claude Code (37%), and Codex (13%). Allsopp notes the figures aren't a direct market-share comparison, because last year's survey asked for a single tool while this year allowed multiple selections.

He also links Claude's position to the platform's SKILL.md files, introduced nearly a year ago, which let users build reproducible processes. 68% of AI-using respondents now use skill files, and 55% create their own — though Allsopp cautions that "things can swiftly change in this era of models one-upping each other."

Trust and quality perception

Two-thirds of marketers (66%) trust AI-generated content only if a human reviews it first. Just 2.7% trust it with no review at all. Distrust splits sharply by usage: only 6% of AI users generally distrust the output, versus 57% of those who don't use AI at all.

Quality perception depends heavily on who you ask. 56% of AI users say human-written content is better quality than their AI content, and 23% call it "way better." But 43% of founders and business owners say their AI content beats human writing, compared to just 17% of junior and mid-level marketers.

More than half of respondents (61% combined) say their view of AI-generated content improved over the past year: 38% somewhat more positive, 23% much more positive.

Editing is the norm — disclosure is not

Growth in AI-assisted content doesn't mean growth in unedited writing. Among respondents who publish no pure AI content at all, 58% still reported their AI-assisted share has grown. Meanwhile, the share of AI users publishing content that is more than 75% pure, unedited AI rose from 4% in 2025 to 7%, while those publishing zero pure AI content fell from 50% to 40%.

Quality checks are evolving too. 29% of AI users now ask AI to review their AI-generated content, up from 14% last year. Manual self-checking remains the most common approach, and fewer respondents this year skip checks entirely.

Disclosure lags behind production. 78% of marketers using AI don't tell website visitors that AI helped create their content; 22% disclose openly, up from 16% in the 2025 survey. Allsopp points out that definitions vary — getting AI feedback on stats or structure differs from having it write most of the page. He also found that those publishing the most pure AI content were no more likely to disclose it.

The money question

Almost half of AI users say the technology hasn't saved them money on writers or freelancers, and 26% don't know whether it has. A clear explanation: only 28% of AI users work with outside writers at all. The rest write content in-house or themselves. Among those who did report savings, the typical range is $1,001–$2,000 per month.

With adoption now at 93% and specialized tasks like data analysis and keyword clustering more than doubling year over year, the open question for search marketers is whether rising volumes of AI-assisted content will trigger measurable ranking effects — and whether the 7% now publishing mostly unedited AI output grows when next year's survey runs.