Why Your Traffic Fell While Your Rankings Didn't
A marketing director sent me a screenshot last month. Google Search Console. 90-day view.
Impressions: up 22%. Clicks: down 18%.
Her first message: "The SEO agency said everything's working."
Her second, an hour later: "Why is everything working and traffic is still going down?"
I've gotten versions of this question a dozen times in the last six months. Different industries, different-sized brands, same chart. The metrics that used to move together stopped moving together — and nobody on the existing team has a clean explanation for why.
Here's the honest one.
The playbook is running on dead assumptions
For twenty years, digital marketing operated on a chain that felt like physics: impressions led to clicks, clicks led to sessions, sessions led to revenue. Optimize each link in the chain and the business grew. The logic was clean. The feedback loops were tight.
That chain broke at the first link.
The SEO discipline was designed for a world where receiving an impression meant a person was about to decide whether to click. That's no longer the world. A growing share of impressions now end inside the AI interface. The user received the answer, closed the tab, moved on. Your rankings held. The SERP still shows your page. But above the first organic result, there's an AI-generated response — and for informational queries, that response is where the journey ends.
The technical name is zero-click search. More than half of online searches now end without a website visit. That share has moved in one direction since 2014. AI accelerated it.
This is not a failure of SEO. Your content is being indexed, your pages are being ranked, the system is working exactly as designed. The design was built for a world where clicking through was the natural next step. That world is smaller every quarter.
What I'm finding across client accounts
Optimizing CTR treats the symptom, not the source.
The instinct when clicks fall is to improve the title tag. Better meta description. A/B test the snippet. Run a few variants.
None of that addresses why the click didn't happen. The click didn't happen because an AI engine answered the question one step earlier in the journey — before the user ever reached the organic results. No headline improvement changes what happens in the layer above your page.
I've watched brands spend three months iterating on title tags while their AI visibility problem compounded untouched. The chart kept getting worse. The titles kept getting better. These are not the same problem.
Informational queries went first. The consideration layer is going next.
Zero-click started with purely informational queries — how does X work, what's the difference between Y and Z, who handles this type of problem. Those moved into AI interfaces and stopped coming out.
Commercial queries — ones where the user intends to buy — still drive clicks, because a user who wants to purchase still needs to visit a page to complete the transaction. That's stable for now.
But watch the middle layer. The research phase — where someone forms a preference before they're ready to buy — is moving fully into AI. Customers are arriving at your site with opinions already formed. If your brand wasn't present in the AI conversation during that research phase, the customers who do arrive may already have a different brand in mind.
Ranked and recommended are now two different things.
Search engines reward pages. Answer engines reward entities.
Those two systems use different criteria, produce different winners, and require different strategies. A brand can hold position one on a high-volume keyword and receive zero AI citations for the same query — because citation is driven by entity clarity, external corroboration, and answer-first content structure. Not by keyword density or link equity alone.
I've seen this pattern consistently: a brand visible at the top of search results, invisible in AI answers for the same queries. The SEO is working. The AEO problem is untouched.
The fix is upstream of your website.
AI forms its opinion about your brand from third-party sources before it ever reaches your domain. Industry publications. Review platforms. Community discussions. Directories. It cross-references these to build its understanding of what you are and whether you're worth recommending.
Most AEO strategies start with schema markup on the brand's own site. That's the last mile. The brands getting cited have clean, consistent entity definitions across those external sources. The brands not getting cited are spending their optimization budget on properties AI hasn't validated them through yet.
You're doing the last mile while ignoring the first twenty-six.
The first question to ask
Before you touch a title tag, open ChatGPT and Perplexity. Type the question your best customer asks when they're actively looking for what you sell — not a keyword-research query, the actual question from your sales calls and DMs.
Does your brand appear?
If it doesn't, that's your baseline and your problem statement. Not a click-through rate problem. A recommendation problem. The fix is entity clarity: a consistent, machine-readable definition of what you are and who you serve, built across the sources AI consults before it reaches your site.
That's different work than SEO. It's also more durable — because it compounds the same way brand authority always has.
Most will adjust the meta descriptions
Most people who read this will go back to their analytics, look at the divergence, and add it to the list of things to monitor. Uncomfortable but not urgent enough to change the strategy.
A few will ask a different question: what does AI actually think about my brand right now?
That's where this work starts. And if you're in that second group, the AEO Changelog is where I document what I'm finding — the patterns across client accounts, the specific mechanics, the things that actually moved the needle.
Next: What Nobody Tells You About Getting Your Brand Into AI Answers — the four things a year of citation research kept showing, and what to do about them.
Let's build.