Ahrefs released the findings of its second annual State of AI in Content Marketing survey on September 30, 2026, with a press release following on October 1. Among marketers who use AI, 7% now say more than three-quarters of what they put out is pure AI material with no editing or additional input, up from 4% a year earlier, even as 89% of AI users manually vet what the tools produce, up from 80%.

In Short

Ahrefs asked about 300 marketers how they use AI to make content, and 93% said they do. More of them now rate AI writing above human writing, yet nearly all still have a person read AI material before it goes live. What changes is the workload: more content is made with AI each year, human review stays in place as the main quality control, and a small group of 7% publishes pure AI text with no editing.

What Ahrefs measured

The survey was fielded in August 2026. According to Ahrefs' press release, it covered 300 marketing and SEO professionals, 68 of them in upper management. The full report, written by Glen Allsopp, Head of Marketing Strategy and Research at Ahrefs, counts 301 responses, of which 280 came from people who use AI to create or help create content. That is 93%, against 87% in the 2025 edition. Among the 6.98% who do not use AI, almost all cited accuracy as a concern, according to the report, and more than half said they generally distrust AI-generated material.

Small gaps separate the two documents. The press release gives 49% of AI users as working "much quicker", the report 47.9%. Image creation is 44% in one and 43% in the other. Where the figures diverge, both are shown below. The report also carries a caveat: last year's survey had more respondents, so Ahrefs limits comparisons to high-level answers rather than granular ones.

The questions centre on content marketing workflows, and the sponsor has a stake in the subject. Ahrefs describes itself as an SEO and AI marketing intelligence platform with an index covering more than 170 trillion pages and AI visibility data from more than 455 million monthly prompts. On August 12, 2026, Ahrefs introduced Letaido, an agent-powered workspace for automating multi-step marketing tasks, which the report names as the successor to a tool called Agent A. Respondents were therefore answering questions posed by a vendor whose products sit close to the practices being measured.

Quality perceptions narrow while oversight grows

Opinion on relative quality has moved. According to Ahrefs, 57% of respondents still say human-written content is higher quality than AI-generated content, down from 65% in 2025. The share saying AI content is better rose from 14% to 24%. The press release summary describes that rise as 14%, though its own body text gives the 2025 and 2026 figures that imply a ten-point change.

The report frames the question differently, asking AI users to compare human writing with their own AI content. In that version, 56% of AI users say human-written content is better, and 23% say it is "way better". Seniority shapes the answer. Almost half of founders and business owners, 43%, say their AI content beats human writing, against 17% of junior and mid-level marketers.

Review habits

Better scores have not reduced review. Ahrefs reports that 89% of AI users manually check AI-generated content, up from 80% in 2025, and 66% trust AI-written text only when a human reviews it first. Just 3% generally trust it without further review (2.7% in the report). Some 43% work with subject-matter experts and 33% use an editor, while fewer than 1% say they do not check quality at all.

"Marketers are clearly becoming more confident in what AI can produce, but that doesn't mean they're handing over the keys," said Ryan Law, Director of Content Marketing at Ahrefs.

One change cuts across that pattern. Twice as many AI users now ask an AI system to assess AI-written material: 29% (28.6% in the report), up from 14% last year. Louise Linehan, a content marketer at Ahrefs quoted in the report, described the limits of the approach after trying it. "Even when an AI quality check flagged issues, I wasn't reassured. It just made me wonder what else AI had got wrong or missed," she said.

Such caution has a documented basis. A survey of 565 US-based digital marketers by NP Digital, reported by PPC Land on February 2, 2026, found that 47.1% encountered AI inaccuracies several times a week, 36.5% admitted that hallucinations had reached public content, and only 23% felt confident using outputs without human review. The Ahrefs data suggest the habit of verification has widened rather than faded.

Volume, speed and the cost question

Speed is the clearest gain. Among marketers using AI, 85% say they produce content faster, with 49% working "much quicker" and 36% "slightly quicker" according to the press release. Only 2.5% say AI has slowed production, per the report. Leadership reports more: 62% of upper management respondents say AI has made production much quicker, 13 percentage points above the total, and adoption among that group stands at 97%.

Volume is harder to pin down, because the figures are self-reported shares rather than counts. Across all respondents, 69% say they publish a larger share of AI content than a year ago, rising to 74% among AI users. Among those who report that none of their published content is pure AI, 58% still say their AI share has grown, which the report reads as growth in AI-assisted work rather than in writing produced without human input. Ahrefs' 2025 research found that AI users published a median of 17 pieces a month against 12 for non-users, a gap of 42%.

At the far end, 7% of AI users say more than 75% of what they publish is pure AI, up from 4% in 2025. The share who say none of their output is pure AI fell to 40% from 50%.

Disclosure and spending

Readers are told little. Some 78% of AI users do not tell website visitors that AI helped create content, and 22% do, up from 16% in 2025. Respondents who publish the most pure AI content were no more likely to disclose it, the report states. The regulatory backdrop differs by region: EU transparency obligations under Article 50 carried an August 2, 2026 deadline, with fines of up to 15 million euros or 3% of worldwide annual turnover, and they cover AI-generated text on matters of public interest published without human editorial review. The survey does not ask where respondents are based or whether they fall under that regime.

Savings are uneven. According to the report, 41% of AI users say the technology has saved nothing on writers or freelancers, and 26% do not know. One explanation offered is that only 28% of AI users work with outside writers at all. Among those who did save money, the typical figure is $1,001 to $2,000 a month.

A separate PPC Land report on WARC and TikTok research published July 14, 2026 found a similar split between quantity and quality: 88% of 400 marketers across the UK, US, Australia and Brazil reported higher creative volume from generative AI, while 45% reported a significant quality improvement.

Where AI is used

Brainstorming (78%), outlines (76%), improving content (76%) and editing or rewriting (70%) remain the most common uses. The largest year-on-year increases came elsewhere. Data analysis jumped from 28% to 60%, image creation from 19% to 44%, and creating content from scratch from 44% to 55%. A fourth category appears under different labels: the press release records topic identification doubling from 23% to 46%, whereas the report lists keyword clustering rising from 23% to 45%. The report notes that several of these were among the least popular uses a year earlier.

Opinions quoted in the report are less uniform than the percentages. Si Quan Ong, an SEO and marketing educator at Ahrefs, accepts the speed gain but not that it is good in itself; faster production without strategy, in his words, "just smells factory". Karen Robinson, VP Marketing at Shorthand, draws a line by task, saying "we're not handing it content that needs personality, like our homepage messaging or our annual engagement report". Mateusz Makosiewicz, a marketing researcher at Ahrefs, offered a competitive reading: "It's Cold War logic: once one side escalates, everyone else has to keep up or lose ground."

Claude tops the tool list, with caveats

Tool choice produced the headline in the press release. Claude was selected by 84% of AI users, followed by ChatGPT at 75% and Gemini at 57%; the report gives 83.2%, 75.7% and 57.9%. Among upper management the order held, at 79%, 73% and 55%. The report adds Claude Cowork at 42%, Claude Code at 37% and Codex at 13%, and says 68% of AI users use skill files, with 55% creating their own. Allsopp writes that the introduction of SKILL.md files almost a year ago seemed to coincide with a shift towards the platform, while acknowledging he "may be in a marketing 'bubble' at times".

The comparison with 2025 is loose. The earlier survey asked which tool respondents used to create content, and the press release recalls ChatGPT as the dominant primary tool, followed by Gemini and Claude. This year's question allowed several answers, so the report states the figures are not a direct market-share comparison. Write-in tools were also added to the totals.

Other usage data sit alongside the survey. Similarweb figures reported on September 25, 2026 put claude.ai at 950 million visits in August 2026, up 540% from 148 million a year earlier, with paid sources at 0.46% of traffic against 3.5% for ChatGPT. Anthropic had opened Claude's automation features to non-developers through Cowork in a research preview on January 12, 2026, at first for Max subscribers on macOS. On the OpenAI side, OpenAI reported 900 million weekly ChatGPT users on February 27, 2026, after 800 million in early October 2025, and the Ahrefs report says the figure passed 1 billion by the end of July. Gemini reached 900 million monthly active users in May 2026, though that is a monthly measure set against ChatGPT's weekly one.

Why the figures matter for the marketing community

Three threads run through the data. The first is where effort has moved. Drafting time has fallen for 85% of AI users, yet 89% still run manual review and a third use an editor, which places verification, rather than production, at the centre of the workflow. Does faster production without matching review capacity leave a gap? The survey cannot say, since it measures habits rather than error rates.

The second is the split between those who commission and those who edit. Founders and owners rate AI content far above human writing more often than junior and mid-level staff do, 43% against 17%, and upper management reports bigger speed gains. Perceptions of quality therefore vary with proximity to the work.

The third is transparency. With 22% disclosing and the regulatory picture in the EU tightening, the gap between publishing practice and labelling practice is the least settled of the findings.

Limits apply. The sample is about 300 people, the answers are self-reported, the 2025 sample was larger, question formats changed for tool use, and the sponsor sells products in the same field. The year-on-year movements are real in the data but modest in statistical weight.

Timeline

Summary

  • Who: Ahrefs, the SEO and AI marketing intelligence platform, surveyed about 300 marketing and SEO professionals, 68 of them in upper management.
  • What: The 2026 State of AI in Content Marketing Survey found 93% of respondents using AI for content, 89% of AI users manually reviewing its work, 24% rating AI content above human writing, 7% of AI users saying more than three-quarters of their publishing is pure AI, 22% disclosing AI use, and Claude selected by 84% of AI users.
  • When: The survey ran in August 2026. Ahrefs published the full report on September 30, 2026 and issued a press release on October 1, 2026.
  • Where: The report appears on the Ahrefs blog at ahrefs.com/blog/marketers-ai-content-creation/, and the press release was distributed from Palo Alto, California. The survey does not state respondents' locations.
  • Why: Ahrefs is tracking, for the second year, how AI changes content marketing practice. The results show rising adoption and speed alongside rising human review, a narrowing perceived quality gap, and limited disclosure to audiences.