Your AI Isn't Failing. The Tactics You Cut Are.
Your publishing calendar is fuller than it was two years ago. Your pipeline is not. In September 2026, Orbit Media published its twelfth annual blogging survey of 1,042 content marketers and found that 92.4% of them now use AI somewhere in their content process, and that adoption, in Orbit's own words, "correlates with nothing."
Most advice about this moment tells you to make the content better, or smarter, or more AI-assisted. My position is the opposite of that. The AI you bought is measurably uncorrelated with your results, and the repair is to re-fund the expensive human work you quietly cut to afford it: expert-led content, original research, and a human editor with a name. AI belongs in the drafting chair. It does not belong in the strategy chair.
The blog didn't stop working. The parts that made it work got defunded.
Just 13.9% of marketers in that survey say their blog delivers strong results. That is a twelve-year low, six full points below the previous worst year, and roughly half the 26% who said the same thing in 2022. If you are spending $4,000 a month on content, the industry-wide odds that spend is working for you just fell to about one in seven.
The second number is worse, and it gets far less attention. Nearly one in five marketers, 18.9%, now answer "I don't know if the blog delivers results." For the previous decade that figure sat between 9% and 14%. So the group who can prove the program works shrank, and the group who cannot tell either way is bigger than it has ever been. You are not imagining the fog. It is measured, it has a size, and it is being paid for monthly.
I have been in search since 2008, seventeen years, and I have watched "SEO is dead" cycles arrive and leave. This one is different, and the reason I can say so is that I run a content operation that adopted AI early and heavily. I built my own tooling in Claude Code sessions, a local content command center I use daily, and I still route every statistic in every post through human fact-checking under a named byline, because the moment a number breaks, the page stops being an asset and becomes a liability. That combination, heavy AI plus protected human steps, is exactly the experiment the Orbit data describes. Which means I can tell you which line items I refuse to automate, and why, starting with the one the survey says matters most: expert collaboration, which fell from 25% of marketers in 2017 to 7% while everyone was buying speed.
There is a separate question about whether blogging is worth it at all in your category, and I answered that one in a post arguing that your industry has already decided whether blogging works. That piece is about the category you are in. This one is about the choice you made inside it. Two firms in the same industry, same budget, same word count, can land on opposite sides of that 13.9%, and the thing separating them is not their AI stack.
AI bought you fifty hours a year and sold you nothing
The average blog post now takes 3 hours 20 minutes to write, down from a 4-hour-10-minute peak in 2022. Orbit does the arithmetic for you: at 57 posts a year and 50 minutes saved each, the typical marketer now banks about 50 hours of writing time annually. Posts also got shorter, 1,312 words on average, down from a 1,427-word peak in 2023. Fifty hours is more than a full work week returned to you. Results fell anyway.
Be careful with the anti-AI conclusion, because the data does not support it. Orbit found no AI use case that correlates with performance, and AI users and non-users are equally likely to report strong results. But non-AI users are more likely to report disappointing results. AI does not raise your odds of winning. It lowers your odds of losing badly. That is a floor, not a ceiling, and a floor is a reasonable thing to buy. The mistake is paying for the floor with the money that was funding the ceiling.
| What AI changed | What it did not change |
|---|---|
| Writing time per post fell from 4:10 to 3:20 | Strong-results reporting fell to 13.9%, a twelve-year low |
| Adoption reached 92.4% of marketers | No AI use case in the survey correlates with performance |
| About 50 hours of writing saved per year | 18.9% now cannot tell whether the blog delivers at all |
| Fewer catastrophic posts (non-AI users report more disappointing results) | The odds of a standout post, which track human tactics instead |
Andy Crestodina, co-founder and CMO of Orbit Media, put the summary in one line in the report: "Marketers just got faster at making the same things." That is the whole trap in one sentence. Speed applied to a mediocre input produces more mediocre output, sooner, at lower cost per unit, with a dashboard that looks busier than last year.
The tactics that predict success are the ones going extinct
Orbit tracks which practices correlate with strong results, and the standout is not a tool. It is other people. Marketers who collaborate with outside experts and influencers are 2.6 times more likely to report strong results, the single best predictor in the survey. Adoption of that practice fell from 25% in 2017 to 7% today. The most reliable lever in content marketing is now used by one marketer in fourteen.
It is not one lonely casualty. Guest blogging dropped 29% in a single year. Influencer collaboration dropped 22%. Original research, which improves the likelihood of strong results by 50%, is published by fewer marketers than before. Formal human editing, where the human-editor group beats the overall benchmark by nearly two to one and AI-only editing scores lowest of every editing approach, is thinning out. Keyword research still separates the winners, and fewer marketers than ever do it.
Read the pattern rather than the individual lines. All six of the results-linked tactics Orbit tracks became less common this year. Every strategy linked to weak performance became more common. That is not a market changing under you. That is an industry systematically defunding the things that worked because the things that worked cost time, and time just got cheap somewhere else on the page.
Handley goes further in the same report, and the continuation is the part worth pinning above your content budget: "Things like publishing original research, human editing, collaboration… These aren't inefficiencies to optimize away with our robot friends. They might look like extra steps. But they're often where the greatest value enters the work."
The stacking numbers make the argument financial. Marketers who use six or more of Orbit's eight results-linked strategies report strong results at 39%, nearly triple the 14% benchmark. Marketers combining four or more of those strategies with six or more hours per post hit 31%. And out of all 1,042 respondents, zero use all eight. Nobody is running the full playbook. Which means the bar for beating almost everyone in your category is lower than your agency has implied.
Re-fund the human line items, in this order
You do not need to fire your AI tools or your agency. You need to move money and hours from volume back into the five things the survey says actually predict outcomes, and you need to be able to say out loud which of them you are paying for this quarter.
Two of these you can do yourself this week without any technical skill. Step two is a calendar invite and a phone recording: call the best operator you know in your field, ask what customers misunderstand, what they would never skip, what changed this year, and what everyone overrates, then quote them by name with their title. Step three is opening your own booking system and counting something true. A physiotherapy clinic that publishes the share of its first appointments that came from people who had already tried something else now owns a number no competitor and no language model can generate. That is the whole mechanism behind the ten percent rule of information gain: the only durable reason to publish is the part a reader cannot get from the other nine results.
Steps one, four and five are where you press your agency. Ask three questions and listen for hesitation. Who is the named human editor on my account, and what is their fact-checking step? Which outside expert did you interview for last month's posts, and can I see the recording? What original data did we publish this quarter that came from my business? An agency running a volume model will answer the first question with a process name instead of a person. That is your answer.
Traffic is the metric that lies to you the longest
Volume metrics are the most watched thing in content marketing and the least predictive. Orbit found 39% of marketers watch traffic-style numbers above everything else, and only 11% of those traffic-watchers report strong results, which is below the 14% overall benchmark. Watching traffic correlates with doing slightly worse than average. Marketers who measure qualified leads, deals, and revenue are more likely to report strong results.
That is why I keep arguing you should stop grading your content on Google traffic. A post that brings 4,000 readers and zero inquiries is a cost center with good optics. A post that brings 90 readers and three booked calls pays a staff salary. Sessions cannot tell those two apart, which is a large part of why 18.9% of marketers can no longer say whether their blog works.
Andrew Holland made the structural point in Search Engine Land on September 14, 2026: "SEO content is getting cheaper to produce, shifting value toward the ideas, expertise, judgment, and experience behind the words." Cheap production is now table stakes, so it cannot be an advantage. Whatever is left after you subtract the cheap part is your entire competitive position.
Frequently Asked Questions
Should I stop using AI to write my blog posts?
No. The Orbit Media 2026 survey found AI users and non-users are equally likely to report strong results, but non-AI users are more likely to report disappointing ones. AI lowers your risk of a bad post without raising your chance of a great one. Keep it for drafting, outlining and cleanup, and put a named human in charge of the facts, the point of view and the final edit, since AI-only editing scored lowest of every editing approach in that survey.
What does expert collaboration look like for a small business?
It is far simpler than the word suggests. Record a 20-minute call with a practitioner your customers would respect, a senior technician, a specialist, a supplier, or a long-tenured member of your own team, then quote them by name and title inside the post. You are not buying a celebrity endorsement, you are adding a first-hand judgment no competitor can copy. Only 7% of marketers do this, down from 25% in 2017, and the ones who do are 2.6 times more likely to report strong results.
How long before cutting my publishing volume shows results?
Plan on two quarters before you judge it, and watch leads rather than traffic in the meantime. Depth compounds slowly because the pieces that earn links, citations and referrals need time to be found and shared. What should change first is the quality of the conversations: prospects start arriving having already read something specific. If nothing changes in six months and you genuinely ran the human tactics rather than just publishing less, the problem is your topic selection, not your cadence.
Go and look at what your content budget bought last quarter. If you can point to the line for expert interviews, original data, or a named editor, you are already running a program most of the 1,042 marketers in that survey are not. If you cannot find those lines, the money did not disappear, it got converted into speed, and speed was never the thing that was working. This was a choice, made quietly, one efficiency at a time. It can be unmade the same way. If you want a second pair of eyes on which of your line items to cut and which to protect, book a call and we will go through them together.
