The Numbers Don't Lie: AI Search Is Eating Traditional Traffic — Here's What Brands Should Do

Two research findings published this week moved the conversation about AI search's impact on traditional traffic from industry consensus to hard evidence.

One, from the University of Washington, quantified AI Overviews' impact on Wikipedia's search referral traffic. The other, from BrightLocal, revealed that the share of consumers using AI tools for local business recommendations surged from 6% to 45% in a single year.

Read those two numbers together and the conclusion is inescapable: AI isn't going to change search someday. It already has.

Two Data Points, One Trend

Let's start with the University of Washington study. Researchers estimated that since Google AI Overviews launched, Wikipedia's search referral traffic dropped by approximately 5%. Google disputes the figure, arguing the actual impact is smaller. But regardless of how both sides ultimately calibrate that number, the direction is clear — AI summaries are indeed siphoning clicks from traditional search results.

The consumer-side shift is even more striking. BrightLocal's 2026 survey found that 45% of consumers now use AI tools for local business recommendations, up from just 6% the prior year. That's not incremental growth — that's a 7.5x jump. ChatGPT was the single most widely used AI tool for this purpose.

For brands, here's what that means: your potential customers are turning to AI instead of search engines to make decisions. And whether your brand appears in AI's answers? You probably have no idea.

"Search Rankings" and "AI Citations" Are Two Different Games

Traditional SEO's core objective: get your page to rank high on search results pages and earn clicks.

GEO (Generative Engine Optimization)'s core objective: get your brand information understood, trusted, and cited by AI models so it appears in AI-generated answers.

The key differences:

Traditional SEO: Battlefield is search results page ranking. Users click links to visit websites. Content relies on keyword matching + backlink authority. Rankings fluctuate significantly. GEO: Battlefield is AI-generated answer content. Users read conclusions AI provides directly. Content relies on structured + credible + sourced information. Compounds over time, increasingly stable.

This doesn't mean SEO is obsolete. Structured data, Schema markup, high-quality original content — these SEO fundamentals are equally critical for GEO. But doing just these isn't enough anymore. You also need your content to be optimized for how AI "reads."

What Kind of Content Does AI Prefer to Cite?

Based on public research and testing across multiple platforms, AI models tend to cite content with these characteristics:

1. Lead with the answer

State the conclusion in your first paragraph. Don't build up to it. AI needs to extract a clear answer from your opening.

2. Structure everything

Tables, lists, and code blocks are far easier for AI to extract and cite than walls of prose. Format comparisons as tables, steps as numbered lists.

3. Use specific data

"Processing speed improved by 32%" is far more citable than "processing speed is fast." Numbers with timestamps, sample sizes, and sources carry more weight.

4. Include rich entities

Content containing specific brand names, product names, technical terms, and authoritative references is more easily recognized by AI knowledge graphs.

5. Maintain cross-platform consistency

When a brand's information is consistent across its website, industry forums, developer platforms, and media coverage, AI assigns higher trust weight.

This Week's Industry Signals

Beyond the University of Washington study, several other developments this week reinforce the same trend:

Google DeepMind published a new Autoregressive Ranking (ARR) model, testing a unified AI model that could replace traditional retrieval and ranking pipelines entirely — meaning the underlying architecture of AI search is still evolving rapidly.

ChatGPT Shopping now leans heavily on product feeds, with product pages earning 24% of AI citations compared to just 4% for Reddit and YouTube.

RankBurn and PallasAI both launched AI visibility tracking tools, specifically designed to help brands monitor and improve their presence in AI-generated answers — a market segment that didn't exist 12 months ago.

Common Crawl analyzed 584,107 llms.txt files, signaling how deeply the infrastructure for AI-readable content is being built out.

What Should Brands Do Right Now?

If you haven't systematically evaluated your brand's visibility in AI search, here are three steps to start with:

Step 1: Run an AI visibility audit

Ask ChatGPT, Perplexity, Google Gemini, and Claude industry-relevant questions about your space. Record whether your brand is mentioned, where it ranks, and whether the description is accurate. No tools needed — you can do this manually.

Step 2: Build structured content assets

Pick 3-5 core topics you most want AI to cite you for. Rewrite or supplement content following the "answer-first + structured + data-backed" principle. Deploy FAQ Schema, HowTo Schema, and other structured markup as priority.

Step 3: Establish ongoing monitoring

AI citations aren't static. They shift with model updates, competitor content changes, and evolving knowledge bases. Set up a weekly or bi-weekly monitoring cadence to track brand mention changes across key queries.

The Bottom Line

Wikipedia's 5% traffic decline and consumers' 7.5x increase in AI recommendation usage are just the tip of the iceberg.

For brands, the window won't stay open forever. The earlier you establish your AI visibility foundation, the better positioned you'll be in AI's knowledge systems.

This isn't a question of "should we do this" — it's "how much longer can we wait?"

---

*Sources: University of Washington AI Overviews traffic impact study (September 2026), BrightLocal 2026 Consumer AI Usage Survey, Search Engine Journal industry coverage, Google DeepMind ARR research paper, RankBurn and PallasAI product launches.*