Pfizer Moved AI Search In-House in 60 Days (the SEO Industry Should Worry)
Pfizer didn't issue a press release about this. There wasn't a keynote or a blog post announcing the strategy shift. The company quietly moved its entire SEO and AI discoverability operation in-house over the course of 60 days, according to Digiday. And that timeline tells you more about the state of AI search than any vendor pitch deck will.
Joshua Palau, VP of Performance Media at Pfizer, said the team "saw the GEO opportunity coming, and recognized how it's reshaping discovery, understanding and trust." That's the polished version. The less polished version: the tools and agencies most brands rely on for search visibility aren't equipped to handle what's happening in AI search right now. So Pfizer built its own capability.
They're not alone, either.
The List of Companies Building Internal AI Search Teams Keeps Growing
U.S. Bank and Georgia Pacific have both stood up internal AI search teams in recent months. Adobe, Hertz, T-Mobile, Lowe's, and Skims are all actively hiring for managerial or directorial SEO and AI positions. According to the Digiday report, eight global clients have brought SEO and AI capabilities in-house in just the last nine months.
Michael Lacorazza, CMO at U.S. Bank, put it bluntly: "Oftentimes, we have more experience than consultants do because we're practitioners living it day to day."
I think he's right, and that quote should sting a little if you work at an agency. The implication is clear. Brands don't trust that outside partners can move fast enough when the landscape changes every few weeks.
The typical structure is a "hub and spoke" model: a central team of five to ten people who own AI search strategy across the organization, with spokes connecting to individual brand or product teams. It's small. It's fast. And it's designed to respond to algorithm shifts in days rather than the weeks or months an agency relationship normally requires.
Why the Urgency Feels Different This Time
The urgency comes from the traffic numbers. Jellyfish analysis found that client web traffic from unbranded searches has decreased 30 to 70 percent. That range is wide, but even the low end should make you uncomfortable if organic search drives any meaningful part of your acquisition.
According to eMarketer, roughly 26.4 percent of the U.S. population will use generative AI for search in 2026, growing 12.7 percent year over year. One in four Americans is now getting answers from ChatGPT, Gemini, or Perplexity instead of (or before) clicking through to your website.
And the part that makes traditional SEO tools mostly useless for this problem: ranking on the first page of Google no longer guarantees you'll show up in AI-generated answers. We wrote recently about how you can rank first on Google for everything and still be invisible to ChatGPT. The overlap between top Google rankings and AI Overview citations has been collapsing since mid-2025.
It reminds me of the early days of social media measurement around 2010, 2011. Every brand knew social mattered but nobody had a reliable way to measure it. The tools were all counting different things. The metrics didn't agree with each other. Some CMOs hired agencies to figure it out. The ones who moved fastest built small internal teams and learned by doing. We're in that same awkward middle period for AI search visibility right now.
The Vendor Gap Nobody Wants to Talk About
A cottage industry of GEO (generative engine optimization) vendors is emerging. Share of Model, built by Brandtech Group, monitors how LLMs like ChatGPT, Gemini, and Claude describe your brand relative to competitors. Peec AI tracks citation patterns and sentiment across AI platforms. Semrush, BrightEdge, and others have bolted AI visibility features onto existing platforms.
But I'd push back on the "just buy a tool" crowd here: none of these solutions seem mature enough for enterprise at the scale Pfizer or U.S. Bank needs. Most of them launched in the last 12 months. The methodologies aren't standardized. There's no consensus on what "AI search share of voice" even means, let alone how to measure it consistently across five or six different LLMs that update their models every few weeks.
That immaturity is the actual reason brands are building in-house. It's not the usual "we can do it cheaper" story. The vendor ecosystem hasn't caught up to the problem yet. If you're Pfizer and your brand's visibility in AI search directly affects how patients find information about treatments, you can't wait for the tools to mature. You build the team and figure it out in real time.
We wrote recently about how AI visibility tools that cost $500 a month can be replicated for about $100 with DIY scripts. That piece was aimed at smaller teams. What Pfizer is doing is the enterprise version of the same instinct: when the market doesn't have what you need, you build it yourself.
What Mid-Market Teams Can Do Without a Six-Figure Headcount
Not every brand has the budget for a five-person internal GEO team. And honestly, most don't need one yet. But the direction is clear, and there are concrete things you can do this week.
First, run a basic AI visibility audit. Ask ChatGPT, Gemini, and Perplexity the five questions your customers ask most about your category. Note whether your brand appears in the responses, what sources get cited, and what the sentiment looks like. This takes about 20 minutes. If your brand doesn't show up for any of them, that's your baseline and your problem statement in one.
Second, track citation sources. The brands winning in AI search tend to get cited from specific types of content: Reddit discussions, expert roundups, original research, and product documentation. Pull your last quarter of content and ask honestly whether any of it is the kind of source an LLM would reach for. If it's all blog posts rewriting the same talking points as everyone else, it probably won't get cited. From what I've seen, original data and genuine expert perspective are what the models seem to favor.
Third, check your entity clarity. LLMs struggle with brands that don't have clean, consistent entity data across the web. If your Wikipedia page, LinkedIn company profile, and Google Business listing all describe you slightly differently, the model will pick whichever description it trusts most. That might not be the one you'd choose.
The 18-Month Window Before the Fortune 500 Catches Up
My prediction: by Q4 2027, at least 60 percent of Fortune 500 companies will have a dedicated AI search visibility role. Right now, maybe 15 to 20 percent do. That gap is where the advantage lives for brands willing to move now.
The brands building these teams today aren't smarter. They're just the ones who noticed that organic search traffic started evaporating and didn't wait for their agency to bring it up in a quarterly review.
Pfizer built a team in 60 days. That's fast for a pharmaceutical company that normally takes months to approve a slide deck. It's also a signal that the urgency is real.
I don't think every brand needs to go full in-house on this. But every marketing team needs someone who understands how LLMs select and cite sources, and who checks whether the brand is showing up in AI answers at least once a month. That person might be your existing SEO lead with some new tools. It might be a consultant. What it shouldn't be is nobody.
The gap between brands monitoring their AI search presence and brands ignoring it is about to get very obvious in the traffic reports. And in most cases, by the time you see it there, you're already months behind.
