The short version: using AI to publish more content is quietly wrecking small business visibility, because Google's AI Overviews now answer the generic questions directly and only reward pages with something a machine could not have written. I found this out the hard way in 2025, dropped my publishing volume, and my traffic fell while my actual leads went up. That's the trade worth making.
What I did wrong for six months
Last year I got sold on the same idea most small business owners get sold on: AI means you can finally publish at the volume the big brands do, without the big brand budget. I went from two blog posts a week to five, all drafted in ChatGPT or Claude from a brief, then edited by me for maybe twenty minutes each before they went live.
For about six weeks it looked like it was working. More posts, more indexed pages, a small bump in impressions in Search Console. Then between March and August 2025, organic sessions on my own blog fell by 38%. Not a slow decline, a proper cliff around week seven. I checked for a penalty, a technical issue, a Core Update. There wasn't one. What had happened was simpler and worse: Google's AI Overviews had started answering the exact questions my new posts were built around, directly in the search results, before anyone clicked through to me.
The posts that survived, the ones that kept their traffic and kept converting, were the older, slower, messier ones with my actual client numbers in them, my actual mistakes, my actual opinions that contradicted the standard advice. The AI-assisted volume posts were the ones that vanished from view first.
The bit nobody selling you AI content tools wants to say
Here's the uncomfortable part. If your content answers a question that has one correct, factual, non-controversial answer, an AI Overview will answer it before your reader ever reaches your page, no matter how well you optimised the post. "How much does an AI consultant cost" is a real example, which is exactly why on my own site that question now lives behind a page built around my own pricing conversations rather than a generic explainer, because generic explainers are the first thing AI Overviews swallow whole.
Volume was never really the advantage AI content tools promised. Distinctiveness is, and distinctiveness is the one thing AI cannot generate for you, because it comes from things that happened to you specifically. Joe Pulizzi built an entire content marketing philosophy on this decades before AI existed, and the business lessons from Joe Pulizzi hold up better now than they did in 2015, not worse. His core argument was always that a content business wins by owning a niche nobody else can credibly own, not by out-publishing everyone in a broad one.
The five-step audit I now run before writing anything
I built this after the traffic drop, and I run it on every piece of content before it goes anywhere near AI drafting.
- Step 1: Google the exact question first. If an AI Overview already answers it in full, in under 60 words, I don't write that post. There's no gap left to fill.
- Step 2: Ask "could I have written this without a real client, a real number, or a real mistake?" If the answer is yes, the post is generic and will get swallowed the same way.
- Step 3: Add one number that only exists because I ran the thing myself. Not an industry stat, a lived one. A conversion rate, a drop-off week, a price someone paid me.
- Step 4: Check whether the format itself is the differentiator. A calculator, quiz, or tool someone interacts with cannot be summarised into a three-line AI Overview the way a paragraph can. This is a big part of why I've leaned hard into interactive calculators to engage and convert visitors rather than more explainer posts.
- Step 5: Decide where else this needs to live besides Google. Email, LinkedIn, a webinar. If a post only exists to rank, it's fragile by design.
What replaced the volume
I cut back from five posts a week to two, sometimes one. Every post now has to pass all five steps above or it doesn't get published, full stop. Traffic on the blog is still down from its 2024 peak, roughly 30% down as I write this. But leads from the site, actual form fills and booked calls, are up 22% over the same period, because the people arriving are arriving for a reason an AI Overview couldn't satisfy for them.
The pattern isn't unique to me. Figma grew a genuine content and community moat not by out-publishing design blogs but by building tools people used and talked about, the kind of asset behind most breakdowns of Figma's marketing strategy. GoPro did the same thing with user-generated footage instead of branded content volume, which is the whole spine of how GoPro built a brand that wins. Neither company won by publishing more. Both won by owning something nobody else could replicate cheaply, which is exactly what generic AI drafting cannot give a small business no matter how good the model gets.
Where AI still earns its place in this
None of this means don't use AI. I still use it constantly, for first drafts, for restructuring a messy transcript into a post, for generating ten headline options I'll reject nine of. The change is what I use it for. I stopped using AI to decide what to write about and started using it purely to speed up the writing of things I'd already decided were worth saying because they came from something real.
The other channel I've pulled back into is the one AI Overviews cannot touch at all: my own email list. An AI summary can eat your search traffic, it cannot eat your inbox. Every audit-passed post now goes to my list first, days before it's indexed, which is closer to how Mailchimp's marketing strategy has always treated owned audience as the asset that survives platform changes, because it isn't rented from Google or anyone else.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
If you're a small business owner reading this while your own traffic charts do something strange, check one thing before you panic: pull up your ten highest-traffic posts from twelve months ago and search the exact title phrase in Google right now. If an AI Overview answers it in full above your listing, that's your answer, not a penalty, not a technical bug. Nobody enjoys hearing it, but the fix isn't a plugin. It's writing fewer things, and making sure the ones you do write couldn't have been written by the thing that's now competing with you for the click.
Frequently asked questions
Does AI-generated content rank worse in Google now?
Not automatically, Google doesn't detect and penalise "AI-written" text as such. What ranks worse is generic content that repeats an answer already available elsewhere, and AI drafting tools make it very easy to produce exactly that kind of content at speed, which is the real problem.
Should small businesses stop publishing blog content in 2026?
No, but they should publish less and make each post carry something an AI Overview can't summarise away, a real number, a real client story, a genuine opinion, or an interactive element like a calculator or tool.
How do I know if AI Overviews are the reason my traffic dropped?
Search the exact title or main keyword of your best-performing older posts and check whether an AI Overview answers the question in full above your listing. If it does on most of your top posts, that's a strong signal, not a coincidence.
What should I use AI for instead, if not deciding what to write?
Use it for speed once you already know the topic is worth covering: drafting from your own notes, restructuring transcripts, generating headline options. Keep the decision of what's worth saying, and the real experience behind it, entirely human.
Related reading: AI Overviews Are Quietly Killing Your Website Traffic. Here's What To Build Instead and Best Hootsuite Alternatives for Scheduling Social Media in 2026.
You can see whether AI recommends you in about a minute: it queries five engines and scores whether they mention you, then audits the page for schema, llms.txt and answer-first structure.
