AI Is Killing a Certain Kind of Blog. (Probably Not Yours)
Somebody asks me this every week now, usually with a slightly panicked edge to it.
If Google just answers everything at the top of the page… is blogging still worth it?
Here’s my honest answer, and it comes in two halves you’re not going to like equally.
One kind of website is in genuine, serious trouble. Not “needs a few tweaks” trouble. The kind where the money just stops.
And it’s probably not the kind you’re running.
The trick is knowing which one you’ve got — because the two look almost identical from the outside, and only one of them survives what’s happening right now.
What’s actually dying
The business model in trouble is what I’d call rented attention.
It works like this.
You write something that ranks for a common question. Someone searches, clicks, lands on your page.
They see an ad. You collect a fraction of a penny. Repeat about fifty thousand times a month and you’ve got an income.
Notice what that whole chain depends on: Google choosing to send you a stranger.
I’m sure you have run into this a lot over the years. No money pages, just ads.
Newsy sites are the main ones you see with this. They also have the square clickbait at the bottom of the page. Example:
Some of these big name sites used to get thousands of visitors a day from someone searching a story, so these are a prime example of AI taking away their clicks.
That was never a promise. It was a loan.
And when Google started answering those questions itself, at the top of the page, with no click required… the loan got called in.
The numbers aren’t gentle. Seer Interactive tracked 3,119 search terms across 42 organizations and 25.1 million impressions, and found click-through on searches with an AI summary fell from 1.76% to 0.61%.
Worth knowing they looked specifically at informational searches — the ones most exposed to AI summaries. So that’s the worst case rather than the average. Other studies land anywhere from 15% to 60% depending what they measure.
Pick your favorite. They all point the same direction, and it isn’t up.
If your income depends on a steady supply of strangers, that supply is drying up.
Which kind of site have you got?
Here’s the diagnostic. It takes about ten seconds.
Ask yourself what you’re actually selling.
| Selling attention | Selling something | |
| How you earn | Ads, sponsored placements | Products, services, affiliate purchases |
| What a visitor is worth | A fraction of a penny | Anything from nothing to a lot |
| What you need | Enormous volume | The right people |
| Who owns the audience | Google does | You do, if you’ve captured emails |
| Effect of losing 60% of traffic | Catastrophic | Survivable, sometimes barely noticeable |
If you’re in the selling attention column, the math is brutal. Every visitor is worth almost nothing, so you need floods of them, and the flood is drying up.
If you’re in the selling something column — you sell a digital product, a course, a service, a book — you need far fewer people. Losing nine casual searchers and keeping the one who buys is survivable. For an ad-funded site, that’s fatal.
Same traffic drop. Completely different outcome.
The part nobody says out loud
Now the bit that genuinely surprised me when I went digging, and that I almost never see mentioned.
Ahrefs looked at 863,000 keywords and 4 million AI Overview links to work out where the cited pages actually rank in normal search.
| Top 10 | 38% — down from 76% a year earlier |
| Positions 11–100 | 31.2% |
| Nowhere in the top 100 | 31.0% |
Read that bottom row twice.
Nearly a third of AI citations go to pages that don’t rank in the top hundred.
Pages that, under the old rules, effectively did not exist. Nobody was ever going to click to page eleven.
Why does that happen?
Google now takes your question, quietly splits it into a handful of smaller ones, runs them all, and cites whoever answers the pieces best.
Your overall ranking stops being the bouncer at the door.
And when you are cited, you get 35% more clicks than sites that aren’t.
So here’s the thing everyone’s too busy panicking to notice: AI search reaches deeper into the web than blue links ever did.
A small site with genuinely first-hand material has a better chance of getting cited than it ever had of ranking.
That’s not a consolation prize.
That’s the actual opportunity, and it’s sitting right there while everyone writes doom posts.
Avoid “Commodity” Content:
Google prioritizes “non-commodity” content that provides unique viewpoints or expertise
AI can rewrite anything that already exists. It can’t invent what happened to you.
So the fix is always the same — put something in the post that only you could have put there.
1. Show your actual experience – “How I Fixed It” Principle”
- Say when you did it and which version of the tool you used
- Use your own screenshots — your real, messy screen, not stock photos. And yes your failures
- Write down what went wrong. “Here’s where it broke and how I fixed it” is the most useful paragraph in most tutorials, and nobody else has yours
2. Use your own numbers – “Information Gain”
- A short post with one number you generated beats a 3,000-word guide that repeats what’s already ranking
- Swap vague for specific — not “it got much faster,” but “it went from 6 seconds to 1.8”
- When you quote someone else’s number, link to the original study, not to a blog quoting it
3. Name your method – “Branded Content”
- “Six things to check when auditing your content” gets skipped. Too Generic.
- “The 6-Step Content Check” gets remembered, quoted, and linked to. Give your work a name if you came up with it
- A name gives people something to refer to. A list doesn’t.
Giving your proprietary process a specific name makes it highly citable, turning a routine checklist into a distinct brand entity in Google’s Knowledge Graph
What I’m doing about it on my own site
Fair’s fair — here’s what this looked like in practice for me, because “build an owned audience” is easy to say and annoying to actually do.
I found two leaks. Not clever ones. Embarrassing ones.
The first: my homepage had a big button offering a free guide.
Lovely button. 🌸😊
It linked straight to the PDF.
Anyone could click it, download the thing, and wander off forever. I never got so much as an email address.
Months of people taking me up on an offer and me learning nothing about any of them.
The second was worse. I had a bonus page for people who’d bought my book — two free PDFs as a thank-you.
Same problem. Download links, no capture.
Everyone who ever claimed a bonus was a stranger I never heard from again.
Both are fixed now. The button opens a signup form instead of the file, and the bonus arrives by email after someone confirms.
Took an afternoon. Exhausting. 😩 But fixed.
At the time it felt like housekeeping. Boring busy work I’d been putting off.
It doesn’t help that these email marketing and marketing automation platforms have the worst, most confusing interface.
If you have ever used one or two, then you know. There will be an auto-response or some other part of the sequence, with no easy way to get there to edit it.
You go in circles, round and round and round. Very frustrating!!
Looking at it now, with the traffic numbers in front of me? It was the single most important defensive thing I could have done.
And here is the truth. I did it with AI.
I connected Claude to an MCP from Mailerlite and it audited my email sequence, found the issues and told me how to fix it.
No lie. It would still be broken if I hadn’t decided to give that a try.
An email list is the only audience you actually own.
No algorithm update touches it. Nobody needs Google’s permission to reach you. It just… stays.
If you’ve got a leak like that, go and fix it before you read another word about AI search.
If you need to, pick an AI and do an audit like I did. Pick your favorite AI and ask it if it can do it and what it needs from you to accomplish it.
The AI tool is also the instruction manual.
Going back through the old stuff
The other thing I’m working through: my own archive is out of date and I hadn’t really admitted it.
I opened one of my older posts recently and had a major oh no moment.
Not wrong, exactly.
Just written for a version of Google that doesn’t behave that way anymore — all tactics, no substance, answering a question that now gets answered in the summary box before anyone reaches me.
So I’m triaging the lot. Here’s the rule I’m using, and you’re welcome to steal it:
Can an AI answer this post’s central question completely, in three sentences?
>
If yes, either add something original — or accept it’s done for.
That’s it. That’s the whole test.
Pick an old post
What “something original” means in practice:
- A screenshot of you actually doing the thing — not a stock image, your messy real screen
- A number you generated yourself — what it cost, how long it took, what your result was
- The part that went wrong — nobody else has your specific failure and it’s the most useful paragraph in most tutorials
- A named framework or checklist you made up, that people can refer to
If you can’t add any of that, updating the post won’t save it. It was always something anyone could have written, like a general overview.
Leave it alone or delete it. Just don’t spend your Saturday polishing it.
So is blogging still worth it?
Yes. But the job description changed, and it’s worth being blunt about how.
The old job: write things that get found, and collect the traffic.
The new job: write things worth coming back to, and hold onto the people who do.
That’s harder. It’s also a much better business, because nobody can switch it off overnight.
Truthfully, we should’ve all been doing this all along. Not depending on Google, but we did. The old Google monster got us all.
Here’s what actually moves now:
- Write what only you could write. Your screenshots, your numbers, your mistakes. This is the one input AI structurally cannot generate — if it could produce it, it wouldn’t be original.
- Capture emails immediately if you are not. Not “someday.” Go and check your own free offers this week and see whether they’re leaking.
- Sell something. Even something small. It changes what a visitor is worth from a fraction of a penny to an actual amount.
- Stop writing definitional posts. “What is this or that” belongs to the summary box now. Let it have them.
- Go and cite your sources. It’s one of the few things with real research behind it, and it costs nothing.
Websites that exist to be found have a rough decade ahead of them.
Websites that exist to be returned to are going to be fine.
Frequently asked questions
Is blogging dead?
No, but low-effort blogging is. If your posts summarize what’s already on page one, an AI does that faster and for free. If your posts contain something that only exists because you did the work, you’re in better shape than you were two years ago.
Should I stop writing how-to posts?
Not at all — how-to content still does well, because people want to see the actual steps and screenshots. What’s dying is definitional content: “what is X”, “how long does Y take”. Those get answered in the summary box.
Do I really need an email list?
If you sell anything, yes. It’s the only audience you own outright. Search traffic is borrowed, social reach is borrowed, but an inbox is an inbox. Start it before you feel ready — even a tiny list beats no list.
Will AI Overviews get worse for traffic?
Probably, and planning for that beats hoping otherwise. The sensible move is to reduce how much of your income depends on a stranger clicking a search result at all.
Should I take the ads off my site?
Not necessarily — ad income that still arrives is still income. Just don’t build anything new on that foundation, and don’t expect it to hold up. Treat it as a nice extra rather than the plan.
Can a brand new blog still get traffic in 2026?
Yes, and possibly more easily than in 2020. Nearly a third of AI citations go to pages that don’t rank in the top 100 — so you no longer need years of authority before anything reaches anyone. You need something genuinely worth citing.
Your turn
Go and look at your own site with the diagnostic from earlier. Are you selling attention, or selling something?
If it’s attention, you’ve got work to do, and I’d start this week rather than 6 months from now.
If it’s something else — a product, a service, a list, a reason for people to come back. Then the panic in your social media feed isn’t really about you. Fix your leaks, write the things only you can write, and get back to work.
Which column did you land in? And have you found a leak in your own setup yet? Tell me in the comments — I’d genuinely like to know I’m not the only one who had a download button quietly betraying them for months.
This is part three of a short series. Part one: [I traced the AI SEO numbers back to their source — and the source no longer exists.](https://katcrews.com/ai-seo-statistics/) Part two: [the six AI prompts I run before publishing anything.](https://katcrews.com/ai-content-audit/)