Over the last year, something on the internet collapsed. The organic search results got buried underneath the AI overview. Reddit threads started sounding the same to farm karma. LinkedIn comments turned into "This is A, but not B" thoughts. New laptops, mobiles and devices ship with AI baked in. Simultaneously, AI overviews, Copilot, Perplexity, Gemini and friends showed up everywhere and promised to answer everything for us.
Question is, everything is going to be noisy and what would be a plausible place for real signals.
Content that is expensive to make up
Proprietary, original data
- annual industry benchmarks
- "state of the category" reports
- usage trends only we can see from our own product data
Checkable facts and concrete outcomes
- Instead of "trusted by leading companies", we can say: "X% reduction in Y metric over Z months for A, B, C customers"
- Name the actual companies (with their permission).
- Presence on platforms we don't control (G2, Capterra, Trustpilot, relevant forums) matters more than our own case study pages.
Two-way human contact
Talking to someone directly like DM, call, Zoom is still the best filter. A bot or a paid shill collapses when you ask specific questions like "What exactly happened in week 6 of the programme?" or "What was the worst part about using this tool?"
Identity-bound public track record
A person who has posted for years under their real name, in a real network, with a career they care about, can't fake their way through everything. Their reputation is at stake.
Closed communities with a cost of entry
Paid Discords, alumni groups, professional Slack spaces. Fakery is expensive here because there's some barrier to entry and members can push back if they smell marketing or spam.
Gated content
Substack crossed 8.4 million paid subscriptions in Q1 2026, a 68% jump from the 5 million milestone. It's a clear indication of this. If AI overviews scrape and summarize everything you publish, and you don't even get the click, it's rational for the best creators to go dark.
Sooner or later, AI systems and search engines will have to pay or partner to bring that quality content back into the index. Or the worst case, what if they become the content creator that educates all of them by actually testing the products with their agents, verifying the product information, doing mass surveys and reading people's psychology. That's the worst case. For now, I choose to be optimistic.
Revealed behaviour, not just stated opinion
Instead of writing "people liked this programme", you should be writing "42% of active users completed the programme, and 31% reported applying at least one workflow change within 30 days."
Instead of a CTA that says "most popular seo tool used by leading platforms" / "transform your workflow", say "the (category) tool that has 2x lead conversions. try for free"
When I think about reviews or case studies now, people don't just care what somebody said. They care what changed in their life or business, and whether I can verify that change independently. Reviews mentioning a specific negative are trusted 2–3x more than purely positive ones as it talks more on how the product impacted their workflow rather than a positive review that says "product is really good".
What happens when the buyer is not human?
There's one more twist that may or may not change the whole marketing game. So far we've assumed a human reads the content and decides. But the direction of many tools is:
"Let your agent do the research and come back with a recommendation."
Here, the agent compares tools or programmes. It checks claims, scans reviews, maybe even messages other agents. It then picks one option or shows you a shortlist.
There is not going to be an awareness stage. Straight into consideration and purchase.
Ensure to,
- Stating specific, checkable facts instead of vague claims.
- Using clear structure and improving page speeds.
- Backing up outcomes with numbers and concrete examples.
An agent doesn't care about your brand adjectives but it cares about whether your claims can be cross-checked.
Using AI for Writing is still in the picture
Given everything above, should we stop using AI to write content? I don't think so. But I do think we need to change how we use it.
If AI is doing 100% of the work like from ideas to structure to examples, the result will usually feel like a skeleton.
A healthier split looks closer to:
- 20% AI - helping with structure, tightening grammar, suggesting variations
- 80% human - doing the research, bringing in real examples and product stories, adding perspective and nuance and making sure the voice feels like an actual person, pitching in the right feature sets and ofc the convincing narrative around.
Because, when you prompt the AI to the research you expect nothing but the duplicated content of the competitor pages which is already ranking. Add your perspectives from research and tighten it with AI. I think, that is the only way through!
There's a catch with the detection side,
Wikipedia's editors keep a page called "Signs of AI writing", a long field guide to the patterns that keep showing up in AI text: certain overused words, the rule of three, the "it's not X, it's Y" contrast, em dashes dropped in where a comma would do. But these patterns like writing in threes and reaching for the em dash were present long before the centuries when ChatGPT existed.
Run Emily Dickinson, or the "life, liberty and the pursuit of happiness" line from the Declaration of Independence, through a strict AI filter and a fair bit of it gets flagged. So a Google update that tried to scrub AI content by punishing these signals would be risky. Which is why Google's stand makes more sense. They're not "against AI". They're against low-value content.
People don't give up on trust that easy
There's one last thing that keeps me from being fully pessimistic.
Humans are stubborn about trust. When content costs nothing, trust becomes the expensive thing. The only way to earn it is to own what a machine can't cheaply reproduce: data you collect yourself, outcomes someone can verify, a name with a reputation behind it.
Marketing can win by owning something that's expensive to fake.