AI Is Killing Your Traffic. Your Reaction Is Making It Worse.

I run an SEO agency. I look at traffic dashboards every single day. And over the past year, I've watched the same thing happen across nearly every client account: organic traffic is dropping. Not crashing. Dropping. Steadily, consistently, like a slow leak you can't find.

The culprit isn't a mystery. It's AI Overviews. Google's little answer boxes that sit on top of everything, give users what they need, and send them on their way without ever clicking a link.

But here's the part that really gets me. The way the industry is responding to this problem is actively making it worse.

The Numbers Don't Lie: AI Is Eating Your Traffic

Let me give you the actual numbers, because this isn't speculation.

Organic click-through rates dropped 61% for queries where AI Overviews appear. From 1.76% to 0.61%. The number one position on Google, the spot everyone fights for, saw its CTR collapse from 7.3% to 2.6% in a single year.

Zero-click searches now account for 58-65% of all searches. On queries with AI Overviews? That number hits 83%. Publisher traffic from Google is down 38% year-over-year as of January 2026.

AI Overviews now show up on 35-45% of all searches. For informational queries, the kind that blogs and content marketing depend on, it's over 60%.

This isn't a blip. This is a structural change in how search works. And everyone felt it.

The Panic Response: llms.txt, GEO, and the Rush to Please the Machines

So what did the industry do? What it always does. It panicked. Then it tried to optimize its way out of the problem.

First came llms.txt. Proposed in late 2024, the idea was simple: create a file on your website that tells AI crawlers what to read. A content whitelist for large language models. "Make it easy for the AI to find your best stuff and maybe it'll cite you." We wrote a full explainer on what llms.txt actually is when it first came out.

The problem? A 2026 audit found that zero major AI bots actually request llms.txt files. Not GPTBot. Not ClaudeBot. Not PerplexityBot. None of them. People are creating files that literally nobody reads.

Then came GEO, Generative Engine Optimization. A whole new discipline. New agencies. New tools. New certifications. Except 47% of brands don't even have a GEO strategy, and the ones that do can't measure whether it works. Only 12% of ChatGPT citations match Google's first-page URLs. Optimizing for one doesn't guarantee the other.

And here's the kicker: 50% of content cited in AI responses is less than 13 weeks old. So even if you do get cited, the treadmill never stops. You have to keep producing, keep publishing, keep feeding the machine.

If you want to see what's happening behind the scenes, check out how to see what ChatGPT actually searches for. It's eye-opening.

The Irony: Optimizing for AI Is Making Your Content Worse

Here's where it gets really painful.

In their rush to be "AI-friendly," companies are restructuring their content for machines instead of people. More structured data. More FAQ schemas. More citation-friendly formatting. Less personality. Less opinion. Less of what actually made the content worth reading in the first place.

The results speak for themselves. In a controlled experiment, AI-generated articles averaged 52 visitors per month. Human-written articles on the same topics? 283 visitors. Five times more.

Consumer preference for AI-written content has dropped from 60% to 26% in three years. People can tell. They don't like it. Merriam-Webster named "AI slop" their Word of the Year for 2025. Mentions of the term increased ninefold compared to 2024.

We covered the broader AI copywriting trends last year. The trajectory hasn't improved.

Everything is starting to sound the same. The same phrasing. The same bland tone. The same robotic rhythm. When you write for machines, you lose exactly what makes content rank in the first place: personality, originality, genuine expertise.

Google explicitly warns that AI-generated content without added value violates their spam policy on scaled content abuse. So the very act of trying to game the AI is producing the kind of generic content that both AI and humans will eventually ignore.

The Content Doom Loop Nobody's Talking About

Here's what I think is actually happening, and nobody seems to be connecting the dots.

AI scrapes your content. It produces an answer. Users don't click through to your site. Your traffic drops. You panic. You use AI to produce more content, faster, cheaper. The content quality drops. AI scrapes that worse content. The AI answers get worse. Google gives fewer clicks to low-quality sites.

Round and round. Everyone loses.

The web is eating itself. And the companies that are moving fastest to "adapt" to AI search are the ones accelerating the spiral. More content, less quality. More optimization, less humanity. More output, less reason for anyone to visit the actual website.

Citation decay makes it even worse. Half of the content cited in AI responses is less than 13 weeks old. So you're not just on a treadmill. You're on a treadmill that speeds up every quarter.

What Actually Works (From Someone Who Does This Every Day)

I'm not going to pretend I have all the answers. I don't. But I do look at traffic data across multiple client accounts every day, and here's what I'm seeing.

Google themselves said that optimizing for AI search "is the same as doing SEO for traditional search." That's not me being naive. That's Google telling you the fundamentals haven't changed.

The data actually supports a silver lining. AI search visitors who DO click through to your site convert at 23 times the rate of traditional search visitors. Brands that get cited in AI Overviews earn 35% more organic clicks and 91% more paid clicks.

The traffic is smaller. But it's better.

The fundamentals haven't changed. If you need a refresher, here's our guide to improving your Google rankings. The playbook still works. What doesn't work is abandoning it for some unproven "GEO strategy" that nobody can measure.

Here's what I tell my clients:

Double down on original insights and first-hand experience. Google's E-E-A-T framework rewards exactly this. AI can't fake having done the work.

Stop chasing AI visibility metrics you can't measure. If you can't track it, you can't optimize for it. Focus on what you can measure.

Focus on conversion, not just traffic volume. If your traffic drops 30% but your conversion rate triples, you're winning.

Build brand recognition so AI cites you. The websites that get cited in AI Overviews aren't gaming the system. They're the ones with genuine authority.

AI-optimized content that reads like a technical manual will tank your engagement metrics. Here's how to reduce your bounce rate the right way.

The Uncomfortable Conclusion

The real threat isn't AI. It's losing your nerve.

It's seeing the traffic graphs decline and deciding the answer is to write for robots instead of people. It's hiring a "GEO consultant" instead of investing in content that has a genuine point of view. It's creating llms.txt files that nobody reads while your actual readers leave because your content has become indistinguishable from everything else on the internet.

The websites that will survive this shift are the ones that give AI nothing to replace them with. Original research. Genuine expertise. A real point of view. The kind of content that makes someone bookmark the page instead of just reading the AI summary.

2026 is becoming the year of anti-AI marketing for a reason. Brands are differentiating by being explicitly human. Not because it's trendy. Because it works.

Stop asking "how do I optimize for AI?" Start asking "what can I create that AI can't?"

If your content strategy is "make it easier for machines to read," you've already lost.

If you're watching your traffic decline and not sure what to do next, let's talk. I can't promise the numbers will go back to what they were. But I can help you build something the algorithms can't replace.