Two data points landed within weeks of each other, and read side by side they tell brands something they'd rather not hear: the more AI content floods the web, the less people trust it, and the better they get at spotting it.
AI content detection isn't a hypothetical concern for content teams anymore. It's a measured, published number, and it's not a flattering one for anyone shipping AI drafts without real editing. On July 31, 2026, Dan Petrovic of the Australian research consultancy DEJAN published results from a public reader test — the AI vs Human test — built on the first 1,636 answers submitted by real participants trying to tell AI-written text apart from human-written text. The headline number: readers got it right 64% of the time. That's well above a coin flip, and it's the part that should worry anyone treating AI drafts as finished copy.
The part of DEJAN's finding that matters most for content strategy isn't the 64% itself — it's the direction it moves. Accuracy improves the longer a reader stays engaged with the content. People don't get worse at spotting AI writing as they read more of it; they get better. That inverts a comfortable assumption a lot of content teams have been running on, which is that a skimming reader is an easy mark. The opposite is true. The readers who matter most to you — the ones who actually finish the piece, the ones deciding whether to trust your brand with a purchase — are exactly the readers most likely to clock that the copy was phoned in.
The AI Content Detection Problem Brands Aren't Watching
Most teams tracking AI content detection are thinking about it from the wrong direction. The conversation has been dominated by detector tools — software that scores a piece of text for "AI-ness" — and whether those tools are accurate enough to trust. That's a real question, but it's the wrong one to spend your attention on. DEJAN's test measured something more direct and more consequential: not whether a machine can flag your content, but whether an actual human reader, unaided, notices. Readers don't run your blog post through a detector. They read three paragraphs, feel something's off, and leave. That instinct is what the 64% figure is capturing, and it's the number that predicts bounce rate, not the number that predicts a plagiarism-checker flag.
This is also why schema markup alone doesn't get you cited and why the technical fixes teams reach for first — structured data, cleaner markup, faster load times — don't move the needle on trust. Those are hygiene factors. They don't touch the thing readers are actually reacting to, which is the writing itself: generic phrasing, no specific point of view, no evidence anyone with real expertise touched the piece. You can have perfect schema on a page a reader still bounces from in twenty seconds because the copy reads like a template.
Why AI Search Trust Is Falling While AI Search Use Rises
The second data point widens the problem past your own site and into the search results themselves. A Fractl survey of 1,008 US consumers, reported by Digital Applied on June 20, 2026, found that AI search usage and AI search trust are now moving in opposite directions. Seventy percent of respondents say they increased their use of AI search over the past year — only 3% decreased. At the same time, the share who find AI search results more helpful than traditional search dropped from 82% in 2025 to 54% in 2026, a 28-point fall in twelve months. The share who describe themselves as AI search skeptics rose from 3% to 17%.
| METRIC | 2025 | 2026 | CHANGE |
|---|---|---|---|
| Find AI search results "more helpful" than traditional search | 82% | 54% | -28 pts |
| Identify as "AI search skeptics" | 3% | 17% | +14 pts |
| Increased AI search use over the past year | — | 70% | n/a |
| Say heavy AI use in brand content lowers their trust in that brand | 20% | 39% | +19 pts |
That last row is the one that should reach content teams directly. It's not asking people how they feel about AI search engines — it's asking how they feel about brands. Thirty-nine percent of consumers now say that heavy AI use in a brand's content reduces their trust in that brand, nearly double the 20% who said so in 2025. People are using AI tools more than ever and trusting the output of those tools less than ever, and that skepticism is transferring directly onto the brands whose content looks machine-written. Usage and trust used to move together. They've decoupled, and the gap between them is where brand reputation is getting spent.
EEAT Signals Are What Separate Trusted Content From Suspected AI Content
None of this is an argument to stop using AI in your content process. It's an argument for being honest about what AI drafts are missing before they publish. The eeat signals Google has talked about for years — experience, expertise, authoritativeness, trust — turn out to be almost exactly the list of things a generic AI draft skips. They're also, not coincidentally, the things DEJAN's readers are picking up on when they correctly flag a piece as machine-written. Detection isn't magic; it's the absence of specificity, ownership, and voice, and people notice absence.
This is also the throughline in our own work on what earns a citation from AI engines. The anatomy of an AI citation breaks down the signals that get a brand quoted by name in an AI answer, and they overlap almost entirely with the signals that make a human reader trust the page enough to stay on it. Getting cited and getting trusted are downstream of the same inputs: specificity, sourcing, and a point of view nobody else is offering. And a citation isn't the finish line either — a citation doesn't guarantee the outcome you think it does if the content behind it reads as generic once a person actually clicks through.
What This Means For Your AI Content Detection Strategy
Run some illustrative math on this, and it's not a comfortable exercise. If a tenth of a site's published content reads as obviously AI-generated to readers — a conservative estimate for a lot of high-volume content programs right now — and 39% of consumers say heavy AI use in a brand's content reduces their trust in that brand, the exposure isn't confined to those individual pages. It's a brand-level tax. A reader who bounces off one templated page doesn't file it as an isolated incident; they generalize it to the whole site. That's illustrative, not a cited study, but it maps onto how trust actually works: a few visibly generic pages can color how a reader reads everything else you publish, including the pages a real writer sweated over.
The programs handling this well aren't the ones that banned AI outright, and they aren't the ones that let AI draft and publish untouched either. They're the ones that restructured the process around where a human adds irreplaceable value: reporting a specific number nobody else has, forming an opinion the AI wouldn't default to, and editing hard enough that the structural tells disappear. That's also the shift behind restructuring content operations for how AI search actually works — treating AI as the first draft of the pipeline, not the output.
The trend line here isn't going to reverse. AI search use is climbing — 70% of consumers say they're using it more than they were a year ago — and reader sensitivity to generic AI writing is climbing right alongside it. Waiting for detection accuracy to plateau, or for AI search skepticism to ease off, is a bet against both curves at once. The brands that come out ahead are the ones treating content marketing as the discipline of adding what a template can't: a name, a number, an opinion. Start the audit this week. The pages losing you trust right now are the ones you can identify in an afternoon.
See where you are cited today
A free snapshot audit of your rankings and AI citations before we ever talk.
Josh leads work at the intersection of SEO and generative engines at Something Inc., helping B2B brands get ranked and cited across every major AI engine.