Last Tuesday I opened SEMrush expecting to see the usual numbers for a client in the project management space. Top 3 for "best tools for remote teams," top 5 for "async collaboration software" — everything looked healthy. Then I pulled Google Analytics. Organic traffic had dropped 40% week-over-week. No algorithm update. No penalty. No lost rankings.
The traffic was vanishing into thin air, except it wasn't thin air. It was AI.
I ran the same queries in ChatGPT and Perplexity. For "best tools for remote teams," our client wasn't mentioned. Not in the top 3, not in the top 10, not in the sources cited. A competitor we'd been outranking for months showed up as the second recommendation, complete with a specific use case description.
This is the part nobody warns you about: your rankings can stay perfectly still while your actual visibility collapses.
The metric that matters now isn't position — it's presence
Traditional SEO taught us to obsess over position. Position 1 gets ~28% of clicks, position 2 gets 19%, and so on. That math still holds for blue-link results. But AI search doesn't have positions. It has citations.
When ChatGPT answers "best tools for remote teams," it doesn't show a list of 10 blue links. It synthesizes a response and cites 2-4 sources. If you're not one of them, you don't exist for that query. It's binary: cited or invisible.
I started tracking this manually. For 50 high-intent keywords in our niche, I checked which ones triggered AI-generated answers in ChatGPT and Perplexity. Then I checked whether our client appeared in those answers.
The gap was staggering. We held page-one rankings for 38 of those 50 keywords. But we appeared in AI-generated responses for only 11. Meanwhile, a competitor with page-one rankings for just 22 keywords showed up in AI responses for 19 of them.
> Generative Engine Optimization (GEO) is the practice of structuring content so AI models can parse, trust, and cite it — distinct from traditional SEO, which optimizes for crawlers and link graphs.
Why our content wasn't getting cited
I audited our top pages against what the AI models were actually citing. The pattern was clear:
1. We wrote for skimmers, not for synthesis.Our articles were structured like every other SEO piece: intro, H2s, bullet points, conclusion. But AI models don't skim — they parse. The pages that got cited had clear, self-contained sections that answered specific questions without requiring context from elsewhere on the page.
2. We buried the specific behind the general.Our "best tools" page said things like "offers robust features for team collaboration." The competitor's page said "supports up to 500 concurrent users with real-time document editing and 12 native integrations." AI models cite specificity. Vagueness gets ignored.
3. We had no structured data for machines.Our schema markup was basic — Article, Breadcrumb, Organization. The pages getting cited had FAQ schema, HowTo schema, and Product schema that gave AI models clean, extractable facts.
I ran a quick test. I took our lowest-performing page in AI results and rewrote one section to include specific numbers, a clear comparison table, and FAQ schema. Within 72 hours, that page started appearing in Perplexity citations for two related long-tail queries. The change wasn't dramatic — maybe 200 words rewritten and some schema added — but the AI visibility shifted immediately.
The tools that actually help here
Most rank trackers are useless for this. They tell you where you stand in a list that fewer people are seeing. You need to know whether AI models are citing you, and for which queries.
I've been using a GEO Audit Tool to check which of our pages are showing up in AI responses across multiple engines. It's the only way I've found to get a baseline — you can't optimize what you can't measure.
For understanding the broader shift, I keep coming back to the breakdown in GEO vs SEO — it lays out exactly where traditional optimization ends and generative optimization begins. The mental model shift is bigger than most people realize.
And when I need to quickly check whether a domain has any AI search presence at all, I use the AI Gravity Checker. It's blunt but fast — good for competitive reconnaissance.
What I'm doing differently now
I stopped treating AI search as a separate channel. It's not "SEO vs. GEO." It's just how search works now, and the line keeps blurring.
My content briefs now include three new requirements:
The client's traffic has stabilized, but the mix has shifted. We're seeing fewer clicks from traditional results and more from AI-cited snippets. The total volume is recovering, but the path users take to get to us has fundamentally changed.
If you're only tracking rankings, you're measuring a world that's disappearing.
FAQ
How do I know if AI search is affecting my traffic?Check your branded search volume in Google Search Console. If it's holding steady but non-branded organic traffic is dropping, AI search is likely intercepting those queries before users ever see the SERP.
Does optimizing for AI search hurt my traditional SEO?No. The practices that help AI models cite you — specificity, clear structure, schema markup — also help traditional crawlers understand your content. It's additive, not either/or.
Which AI search engines should I care about right now?ChatGPT and Perplexity are the main drivers of referral traffic as of mid-2025. Google's AI Overview is growing fast but harder to track because it sits on top of traditional results. Start with the ones you can actually measure.
How often do AI search results change?Much faster than traditional SERPs. I've seen citations appear and disappear within 48 hours based on content freshness signals and model updates. This isn't a set-it-and-forget-it channel.