AI “slop” is invisible now. And not just to people.

A few months ago we wrote our thoughts on The Race to Mediocracy: What Happens When AI Does Our Thinking For Us? We also admitted we used AI to help write it.

Apparently a lot of you were also feeling the drag of effortless competence, as it became one of our most clicked LinkedIn posts to date.

So here’s the sequel. And like most sequels, it’s louder, a bit more chaotic, and someone does something reckless in the second act.

We were worried about people tuning out.

We should have been worried about the machines too.

People are now asking ChatGPT, Perplexity, Gemini and Google’s AI answers the questions they used to type into a search bar. Those tools give you one answer, built from a handful of sources, instead of ten blue links.

Picture ten businesses publishing the same AI-generated slop on the same topic. They have the same headline structure, the same safe, cliché structure and language and the same five tips. An answer engine doesn’t need ten copies of that. In fact, it doesn’t need or want any of them.

Sameness isn’t a creativity problem anymore. It’s a visibility problem.

If a competitor could have published your article word for word, an answer engine has no reason to pick yours. The content that gets cited is the content with something only you could have said in it.

That’s the theory. Here’s what happened when we tested it on ourselves.

We stress-tested it on our own brand (and our own nerves)

Earlier this year we started an AEO experiment on BLOOM and documented it publicly. When we started, our AI visibility was 0.21% (a rounding error, surely?). Today it’s 21.7%, or “out of the red” as we say in the HubSpot AEO Hub world.

A nice number for us to do a little tail wag in our chairs. But the more interesting part is how that AI visibility number survived.

In September, Google rolled out a spam update. Updates like that exist to flush out low value content, and they can knock perfectly decent sites sideways.

In the same month, we (a HubSpot Gold Partner) rebuilt our entire website and moved it off HubSpot CMS. (Chill, please. HubSpot still runs our CRM, marketing, forms and reporting behind the scenes.) Only the website layer moved, because some of the design we wanted was slow to build natively and we got sick of looking at it.

And then our Director broke our own rules and pushed the new site live without our usual pre-launch checks.

To be very clear: we would never, ever do that to a client. Lucy did it to us, on purpose, to see what would happen. Consider BLOOM the crash test dummy so you don’t have to be.

A platform migration, a Google update and a launch by a hooligan in the same month. That’s the kind of combination that gives SEO people a twitch.

Our visibility held. Then it went up.

The irony, part two

Part one confessed we used AI to write about AI making everything sound the same. Part two has a fresh confession. Our new website was built with Claude Code, an AI coding tool, alongside an experienced developer who is now part of our team.

We don’t think that’s a contradiction. AI was never the problem: unchecked AI is. A tool that helps a developer ship in weeks instead of months is brilliant. A tool that writes your opinions for you, and nobody reads them before they go live, is how you end up sounding like everyone else.

(Yes, we see the parallel with the unchecked website launch. Do as we say, not as we do.)

What we actually did

We can’t tell you exactly which change moved the number. Anyone who tells you they can isolate that is selling something. Here’s where we put the effort:

1. We sound like BLOOM. We built a proper tone of voice and stuck to it. If it reads like the average of the internet, it doesn’t go out.
2. We answered real questions, quickly. Our AEO content is built around what businesses actually ask, with the answer up top, not after four paragraphs of throat clearing.
3. We spoke machine. Schema markup tells answer engines who we are, what we do and where: New Zealand, Australia and the United States. We rebuilt ours properly during the migration instead of hoping it would survive.
4. We picked a fight we could win. We’re not trying to be a “global digital agency”. Specific beats vague, for people and for AI.
5. We showed our working. Doing this in public means our content comes from things that actually happened to us, including the slightly embarrassing bits. You’re reading one of them.

The slop test

Before you publish anything, AI-assisted or not, ask three questions:

1. Could a competitor have published this word for word?
2. Does it say something we actually believe, even if some people disagree?
3. Is there anything in here only we could know?

If you got yes, no and no, congratulations. You’ve successfully made AI slop. Your readers will skim it and the robots will skip it too.

Average is easy to generate. It’s also easy to ignore.

FAQs (yes, we know)

We’re about to do the most predictable thing in AEO: add an FAQ section to the bottom of a blog. In our defense, it works. Answer engines love a clear question with a clear answer. So here are ours, written by humans, checked by humans, and only slightly self-aware.

What is AI visibility?

When someone “GPTs it”, do you show up? That’s it. It’s how often your brand appears in AI answers to the questions that matter to your business.

Is showing up the same as being recommended?

No, and this is the one that matters. Being mentioned in a list of ten is nice. Being the name the AI actually suggests is the business. Visibility gets you in the room. Recommendation gets you the work.

Does using AI to write content hurt your visibility?

Using AI, no. Publishing regurgitated garbage, yes. If your content says nothing a competitor couldn’t, there’s no reason for an answer engine to cite it. Fail.

Can a website migration tank your AI visibility?

It can. Structure and technical signals change. URLs don't have to. Rebuilding our schema properly and keeping our URLs helped ours hold. Want the BLOOM migration without the BLOOM launch process? That’s exactly what we do for clients.

Did you really just add an FAQ to a blog about sameness?

Yeah we did. The difference is what’s in the answers.