AI-written content and SEO in 2026: where we draw the line

Plenty of businesses have already used AI to save time on content before ever bringing in an agency, and now they're leaning on us to make sense of it. So much of the web today feels mass-produced, with content that feels assembled rather than written. So let's answer it properly, rather than leaving you to guess or read between the lines of a services page.

Every word of client-facing website copy that comes out of Abstrakt is written by a human, with the dedicated time and effort needed to understand your brand, goals and the audience(s) you serve. We aren’t being precious, and this isn’t a line we reach for because it sounds nice on paper. It's how we hold ourselves accountable for what we put in front of you.

We're not anti-AI. That would be an interesting take from an agency dedicated to being at the forefront of today's web evolution. We just don't let it write your branded copy, and there's a real distinction worth spelling out.

AI certainly earns its keep on the research side: competitor analysis, digging into a topic, pulling data, knocking out a rough outline or proofing content, so a writer isn't staring at a page for an hour. This is what we like to call a creative spark. It gives us enough momentum to get the ball rolling and toss in the full weight of our expertise.

Writing the copy is a completely different job. That's where someone has to make hundreds of small judgement calls about tone, emphasis and what actually matters to your readers. AI makes statistically likely guesses. It doesn't factor in hours of brand meetings, and most importantly, it can't replace the true voice of your core audience.

This isn’t just our opinion. We could stop at “this is what we believe” and leave it there, but Google has been fairly explicit about where it stands. Their position, going back to 2023 and reaffirmed since, is that AI-generated content isn’t penalised for being used. AI content is usually struck down for being thin: produced at scale, with no real oversight and simply chasing rankings. What Google actually rewards is E-E-A-T (experience, expertise, authoritativeness, trustworthiness), and while AI can mimic these qualities to a degree, it can’t truly have them. True E-E-A-T comes from a person or group of people who know exactly what they’re talking about. It leans on years of experience and carries an emotional undertone, sharpened by research, consultation and the one thing that really fine-tunes expertise: learning from human error.

Google’s 2026 core updates made a bold statement. In what many SEOs have described as a culling exercise, sites heavily front-loaded with AI content lost serious ground, with some reporting sharp drops in traffic after a single update. Meanwhile, sites with a real point of view and a named person standing behind the content held up or grew. That’s not a coincidence. The shift hasn’t been to ban AI. The narrative is changing to “bring back valuable, intent-led content”.

Before writing a single word, we thoroughly analyse your existing site, brand guidelines and tone-of-voice materials. Research informs and guides our strategy at every step, rather than simply being copied over. As a result, our end-to-end production is crafted by a person who can articulate the reasoning behind it.

We measure our success by a straightforward standard: could this text apply to any competitor in your industry, or is it unmistakably tailored to your brand's needs?

This thought process is what sets Abstrakt apart. With a strong quality control system in place, we don't try to compete on price. Quality and meticulous attention to detail are the values we serve consistently to our partners. We can't match a process that spits out pages in seconds, and the great thing is, we aren't trying to. Good thing too, because relying on that kind of process is about to get a lot harder to get away with.

The sheer volume of AI-written content has taken search by storm over the past couple of years. That boom is exactly what's pushed regulators to step in with safeguards. The EU recently introduced rules requiring AI-generated content to be identifiable. Anthropic, the company behind Claude, has responded by building invisible watermarks into everything their model writes. These are subtle, consistent patterns in word choice. No reader would ever notice them, but specialist tools can pick them up once there's enough text to examine.

We knew measures like this were coming. As the push for transparency around AI content grew, it was only a matter of time before more formal checks followed.

What that means going forward is fairly simple. AI-written content is becoming traceable and a light edit won't necessarily be enough to remove the mark. Sites leaning heavily on AI copy are exposed in a way they weren't a year ago, both to Google and to anyone else who goes looking. None of this is foolproof and access to detection tools is still limited rather than public. But the direction it's heading in is hard to miss.

Google has already been moving in this direction on its own. Its spam updates have specifically targeted mass-produced AI content and content built to game rankings rather than help anyone. Sites that leaned on this approach have seen real drops in visibility. In their place, Google keeps pushing forward content with a genuine expert behind it, something written by a person who actually understands the subject and has something real to say. Watermarking just adds another layer to what Google is already doing manually through these updates.