AI Shift in SEO

TL;DR

As search shifted from "ten blue links" toward answers generated by ChatGPT, Perplexity, and Google AI Overviews, the question our customers started asking changed too: not "where do I rank on Google?" but "do the models mention me at all?" I launched an AI Search (AEO) data product to answer exactly that — visibility, citations, and brand mentions across LLMs, delivered as structured data over the API. It landed fast: in its first ~2 months, roughly 20% of customers drove about half of all new revenue.

Context

SEO was not dying — it was fragmenting. A growing share of searches were ending inside an LLM answer where no link gets clicked. Our customers (agencies, in-house SEO teams, and increasingly AI-native tools) needed to know how brands surface inside those answers, and they needed it as data they could pipe into their own dashboards and agents, not just a screenshot in a UI.

Problem

The obvious move was to bolt "AI" messaging onto the existing SEO API. I decided against it. The buyer was different, the data was new, and the use cases (monitoring, agent grounding, competitive AEO tracking) didn't map onto our keyword-and-backlink heritage. So I treated this as its own product line with its own positioning, not a relabel.

The Insight

When I looked at where the early revenue actually came from, the distribution surprised me. I had quietly assumed adoption would spread across our base — a little extra spend from a lot of accounts. It didn't. About 20% of customers were responsible for roughly 50% of the new revenue — and a striking share of that demand came from a new kind of buyer: teams building AI products on top of our data, rather than our legacy SEO accounts.

That reframed how I thought about the product. I had been positioning it as a feature for SEO teams. The market was telling me it was infrastructure for AI builders. So I leaned into that: prioritized throughput, clean structured responses, and reliability over the long tail of UI niceties, because the heavy buyers cared about feeding pipelines, not browsing reports. The concentration of revenue wasn't a risk to smooth over — it was the signal pointing at the real wedge.

What I Did

I scoped and shipped an AI Search data product exposing AEO signals over the API: which prompts surface a brand, which LLMs cite it, share of voice against competitors, sentiment, and the source pages models pull from. The point was to make AI visibility a first-class, queryable dataset — the same way rank tracking made Google positions queryable a decade earlier.

• AEO visibility and citation data across major LLMs and AI Overviews, returned as structured JSON built for ingestion, not eyeballing.
• Competitive share-of-voice and brand-mention tracking so customers could benchmark, not just self-monitor.
• API-first delivery so AI tools and agents could ground their own products on our data rather than scraping models themselves.

Key Decisions

Keeping this distinct from the core API let me move on its own clock: position it for AI builders rather than SEO managers, prioritize the data shape those builders needed, and avoid diluting the existing API's roadmap. A relabel would have forced the AEO data to inherit assumptions that no longer fit the buyer.

Results

Lessons Learned

1. New buyers need new products, not new labels. Serving AI builders through an SEO-shaped product would have buried the signal. Separating it surfaced who was actually pulling.
2. Read the revenue distribution, not just the total. The shape of where money came from told me the real use case faster than any roadmap survey would have.
3. Ship into the shift, not after it. Treating AI Search as a queryable dataset early — while everyone else was still arguing whether SEO was dead — is what let a brand-new line carry half of new revenue in its first two months.