AI has made producing content almost effortless. It hasn’t made deciding what to believe any easier. The next phase of the internet could belong to the people who can prove what they say.
The short version
- Producing information is close to free. Human attention isn’t. The bottleneck has moved from making content to finding, judging and checking it.
- AI slop is cheap credibility: pages that look researched without containing any research.
- Trust is already slipping. In a Gartner survey published in September 2026, 57% of US consumers said AI-generated marketing had made them less trusting of brand messaging.
- The next edge is proof: first-hand experience, original data, open methods, a named author and corrections left in public.
- Publish less that says nothing. Publish more that demonstrates something.
The question nobody has answered
There is a question I don’t think the AI industry has properly answered yet: what happens after everyone has access to AI?
I don’t mean when businesses adopt it, or when Google adds another AI box above the results, or when another million sites publish articles written in 30 seconds. I mean afterwards.
We’re approaching a strange point in the internet’s history. Producing information is becoming almost free. Producing something worth paying attention to isn’t. The gap between those two things is going to matter a lot.
The internet has a supply problem
For most of the web’s history, publishing was limited by the time, money and expertise it took. Writing an article took time. Research took longer. A video, a website or a useful resource needed skill and effort.
AI has changed that. One person can now generate hundreds of articles, thousands of product descriptions, and a stream of social posts, images and videos at a scale that used to need a team.
The incentive is obvious. If you can produce more pages for less money, why wouldn’t you?
Here is the problem. The supply of content can grow almost without limit. The supply of human attention can’t. We still have the same hours in a day, and we still have to choose which sites to visit, which articles to read and which people to believe.
The bottleneck has moved. It’s no longer the production of information. It’s the discovery, evaluation and verification of it.
AI slop is a symptom
It’s tempting to treat AI slop as a temporary problem. Better models arrive, writing improves, the internet settles down. I don’t think that’s the whole picture.
The issue is that AI can produce convincing content without any underlying experience, evidence or accountability. An article can look researched without containing original research. A product review can sound authoritative without the writer having used the product. A financial explainer can be beautifully structured and still miss the context that makes it useful.
The surface gets better. The substance doesn’t automatically follow.
Researchers writing in ACM AI Letters in March 2026 describe prototypical AI slop by three features: superficial competence, asymmetric effort and mass producibility. (Their paper also argues slop has cultural value worth studying, which is a fairer reading than most commentary gives it.)
Superficial competence is the part that matters for search. The cost of making something look credible has collapsed. When credibility becomes cheap to manufacture, people need other ways to establish it.
The trust problem is already visible
A Gartner survey of 1,006 US consumers, published in September 2026, found 65% believed brands were producing too much AI-generated marketing content. 57% said it had made them less trusting of brand messaging.
A separate Gartner survey, published in June 2026, found 49% of US consumers thought generative AI had made the quality of available content worse.
In the UK, a DoubleVerify and YouGov study of 2,000 consumers found 92% said the growth of AI-generated content was making it harder to trust information online.
None of this shows people abandoning AI. It shows them getting more selective about what they consume.
For anyone in SEO, publishing or digital marketing, that changes the job. Appearing in more places isn’t enough. The aim is to become a source that people, and increasingly machines, have a reason to trust.
AI has made the production of information cheap. It hasn’t made trust cheap.
The next competitive advantage is proof
If I can generate an article about almost anything with an AI tool, so can everyone else.
What I can’t generate is your experience. I can’t manufacture ten years in an industry. I can’t retrospectively create the dataset from an experiment you ran last month. I can’t honestly claim to have tested a product, interviewed a customer or witnessed an event that never happened.
Take two articles. The first explains how AI search is changing SEO, with familiar observations, industry statistics and a list of recommendations. The second measures how four AI engines answer the same question, publishes the prompts, records the date, shares the raw results and says what the test can’t establish.
Both might be well written. Only one gives the reader something they can inspect for themselves. That is a different kind of value, and it’s much harder to reproduce at scale.
| What AI makes cheap | What it can’t make for you |
|---|---|
| Writing an explainer | Having done the thing |
| Summarising other people’s statistics | A dataset nobody else collected |
| Sounding authoritative | A named person accountable for the claim |
| Publishing on a schedule | Correcting the record in public |
Search will have a verification problem too
AI search engines have to select information, synthesise it and present an answer. They also have to decide which sources deserve to be included. As generated content piles up, that selection gets harder and matters more.
A page that repeats what hundreds of other sites say gives an answer engine little to work with. A page with an original observation, a useful dataset, a clear method or a first-hand account gives it something distinctive.
Original research doesn’t guarantee an AI citation. Publishing a CSV doesn’t create authority on its own. Distinctive, verifiable information simply gives readers and machines a stronger reason to pay attention.
There is a difference between being another source of an answer and being the source of something that wasn’t available before.
The middle of the market gets squeezed
I suspect the biggest casualties of the post-slop internet will be the businesses and publishers in the middle. The exceptional experts will be fine. So, mostly, will the giant publishers with established audiences.
The vulnerable group is the sites whose main advantage is producing reasonably competent information at reasonable scale. That model worked while content was expensive enough to be a barrier to entry. AI has worn the barrier down.
If your site exists to publish another explanation of something thousands of sites have already explained, what is your advantage? Being slightly cheaper or slightly faster won’t last.
Explanatory content still has a place. It needs a reason to exist beyond someone having searched for a keyword.
What comes after the slop?
I think we move towards an internet where provenance, reputation and verification are worth more. Not because every reader becomes a fact-checker, but because the cost of getting information wrong becomes more obvious.
Five things gain value:
1. First-hand experience
Showing that someone has actually done the thing they’re writing about.
2. Original data
Findings that can’t be recreated by asking a chatbot the same question.
3. Transparent methodology
Where the information came from, when it was collected and what its limits are.
4. Recognisable expertise
Knowing who is behind a piece of information and why their view deserves attention.
5. Accountability
Revisiting claims, correcting errors and showing when predictions didn’t come true.
None of this is new. Good journalism, science and serious publishing have relied on it for decades. What’s changing is its relative value. When producing an acceptable article is trivial, producing one that survives scrutiny is the differentiator.
How I try to do this here
Everything on this site is meant to be checkable. A few examples:
- AI search statistics, traced to the source: each figure followed back to where it started.
- UK news robots.txt, November 2026: a dated monthly snapshot of which UK news sites block AI crawlers.
- Predictions: calls dated and scored in public, with wrong ones left visible.
- How to check a statistic before you use it: five questions to ask of any number.
- Method: how the research here is done.
The opportunity isn’t to stop using AI
Rejecting AI isn’t the answer, and people were perfectly capable of producing rubbish before it arrived.
AI is useful. It speeds up research, organises information, makes hard subjects easier to explain and removes repetitive work. The mistake is confusing the ability to generate something with the ability to show it’s worth publishing.
AI should make it easier to do the work behind good content. It shouldn’t become an excuse to skip that work. The sites that gain most will use AI to improve their research, analysis and presentation while investing more in what can’t be generated on demand.
The next internet belongs to people who can prove it
We spent years competing to publish more content, target more keywords and capture more traffic. Then AI made publishing dramatically cheaper, and the obvious reaction was to produce even more.
The more interesting move is the opposite one. Show the data. Explain the method. Put a date on the finding. Link to the original source. Leave the mistakes visible. Give people a reason to come back.
The internet doesn’t need another million articles explaining what AI is. It needs more people showing what they’ve found, what they’ve tested and what they can actually prove.
AI has made the production of information cheap. It hasn’t made trust cheap. That may be the most important distinction in the next phase of search.
Questions people ask
What is AI slop?
In “Why Slop Matters” (ACM AI Letters, March 2026), Kommers and colleagues describe prototypical AI slop as content with a surface of quality but little substance, made with far less effort than it would take without AI, and produced for mass distribution.
Does original research guarantee an AI citation?
No. It gives readers and answer engines something distinctive to check and attribute. Whether a given engine cites it, and credits it correctly, is a separate question. I’ve set out how I’m testing that in this study plan.
Are consumers rejecting AI content?
Not outright. The surveys above point to selectivity rather than rejection: people say AI content makes trust harder, and Gartner reports consumers getting choosier about which sources they rely on.
Cite this post
Whitmore, J. (2026) ‘What happens after the AI slop?’, jacobwhitmore.com, 4 November. Available at: https://jacobwhitmore.com/what-happens-after-ai-slop/
Free to quote with a link back.
Sources
Opened on 5 November 2026.
- Kommers et al.: Why Slop Matters, ACM AI Letters 1(1), 26 Mar 2026
- Gartner: Marketing survey finds 35% of consumers rely on influencers less due to AI (1,006 US consumers, surveyed May to June 2026), 22 Sep 2026
- Gartner: Survey finds 49% of US consumers say GenAI has made content quality worse (307 US consumers, surveyed March 2026), 9 Jun 2026
- DoubleVerify: DoubleVerify/YouGov study on UK consumers and AI-generated content (2,000 UK consumers), 9 Jul 2026