My Traffic Was Dropping and Every SEO Tool I Had Told Me Everything Was Fine
For about four months I watched a slow, unexplained traffic decline on a content site I run alongside my main product.
Not a dramatic drop. Nothing that triggered any alerts. Just a quiet, consistent downward trend on pages that had been stable for over two years. Keyword rankings unchanged. No manual penalties. No technical issues I could identify. Domain authority fine. Backlink profile healthy.
Every tool I had pointed at the site told me the same thing: nothing is wrong here.
But something clearly was. And it took me an embarrassingly long time to figure out what.
The gap in my toolset I had not noticed
The answer came from a conversation in a founder Slack group I am in. Someone asked whether anyone was tracking their AI search visibility — specifically, whether their brand was appearing in ChatGPT, Perplexity, and Google's AI Overviews when users asked questions in their niche.
I had not thought about it at all. My SEO stack was built entirely around traditional search — rankings, backlinks, crawl health, page speed. All of it oriented toward a world where the goal was to appear in a list of blue links on a results page.
The problem is that an increasingly large share of informational queries never reach a results page anymore. They get answered directly by an AI model. And whether your content gets cited in those answers has almost nothing to do with where you rank in traditional search. It is a different signal set entirely.
My rankings were fine. My AI citation visibility — how often my content appeared in AI-generated answers for relevant queries — was apparently not fine. And I had no way of measuring it with any tool I currently owned.
What I found when I started looking
The AI search visibility market has moved fast over the past eighteen months. A range of platforms now exist specifically to track LLM citation tracking, AI Share of Voice, and what is increasingly called Answer Engine Optimisation — the practice of optimising content for AI-generated responses rather than traditional SERP positions.
aiseoradar.com was the platform I ended up using. The positioning was specific and made immediate sense to me: an AI search visibility platform built explicitly for the shift from link-based search to generative AI answers. Not a traditional SEO tool with an AI monitoring layer bolted on — a platform built from the ground up around the way search actually works in 2026.
The core concept the platform is built around is AI Share of Voice — essentially a measure of how often your brand or content appears in AI-generated answers compared to your competitors, across the queries that matter to your business. Not your ranking position on a results page. Not your domain authority score. Your actual presence in the conversations your potential customers are having with AI assistants.
That reframe was genuinely clarifying. I had been asking "why is my traffic dropping" when the more useful question was "is my brand visible where my audience is now searching." The answer to the second question explained the first entirely.
The feature that changed how I think about content
The LLM Citation Gap analysis was the thing that made the platform immediately actionable rather than just informative.
Most visibility tools tell you where you are appearing. This one also surfaces specifically where you should be appearing but are not — the queries where your content is logically relevant, where competitors are getting cited, and where you are invisible. That is a fundamentally different and more useful problem statement than "your traffic is down."
Knowing the gap means knowing what to fix. Without it you are producing content based on keyword research and hoping it gets picked up. With it you have a concrete list of citation opportunities your existing content is missing and a clearer picture of what structural or content changes would close those gaps.
The 5-Bot AI Indexing Pipeline is what makes the data reliable rather than illustrative. Rather than querying a single model on a single day and treating the result as representative, the platform validates content visibility across multiple frontier models simultaneously. AI models are non-deterministic — the same query can produce different answers on different days. A multi-model approach gives you a signal that actually generalises rather than a snapshot that might not reflect typical behaviour.
What the audit actually showed me
Running my site through the platform revealed three things I could not have found any other way.
First, several competitor pieces were appearing in AI responses for queries I ranked for in traditional search. My content was technically performing — it was on the first page of Google — but it was being bypassed at the AI layer in favour of content that was structured more accessibly for language models. My ranking was unchanged. My click-through from those queries had dropped because users were getting answers before they reached the results page.
Second, the Security-First SEO layer the platform includes flagged some technical patterns on the site that were affecting how AI crawlers interpreted my content's authority signals — a category of issue I had genuinely not been aware of before and that does not appear in any traditional audit tool I use.
Third, and most concretely useful: a specific list of topic areas where I had no content competing at the AI citation layer at all — LLM Citation Gaps I had not known existed because they were invisible to keyword-based research. Those gaps represented the queries where my potential audience was getting answers from my competitors instead of me.
The broader shift worth understanding
The reason I am writing this here rather than just fixing the issues and moving on is that I think a lot of people building content-driven products are going to hit a version of this moment.
The generative search transition is not a future concern. It is already affecting traffic for content businesses right now. The percentage of informational queries that get answered by AI Overviews and similar features has grown significantly in the past twelve months. For many categories it is already large enough to show up clearly in analytics without any other explanation for the change.
Traditional SEO ranking tools were built for a world where every search ended with a click to a results page. That world still exists but it is contracting. The tooling needed to measure and act on AI search ranking and generative engine optimisation is different from the tooling built for traditional search — and the gap between what your rank tracker shows and what your actual traffic looks like is going to keep widening as AI-generated answers get more capable.
aiseoradar.com is built specifically for the version of search that exists in 2026. For anyone running a content business where organic discovery matters, understanding where you stand on the AI visibility layer is no longer optional.
The four months I spent confused about why my traffic was dropping while all my traditional metrics looked fine was four months of not knowing there was a different set of metrics I should have been watching.
Link: aiseoradar.com
Has anyone else noticed a disconnect between traditional SEO metrics and actual traffic over the past six to twelve months? Curious whether others are tracking AI visibility yet or still figuring out where to start.
