Why AI Models Don’t Recommend Your Brand

In this week’s episode of Voices of Search, we spoke with Jason Lax, senior SEO and AI expert at SAP. Jason has spent nearly two decades building SAP’s global search program, and today the company is one of the most cited enterprise brands across AI search.

Our conversation covered why rankings no longer tell the full story of a brand’s visibility, why asking a model directly why it didn’t recommend you has become one of the most useful audits available, and why assuming your audience uses the same AI tools you do is a strategic blind spot.

Key Takeaways From This Episode:

• Rankings still matter, but they no longer capture whether a brand is actually being recommended in AI-driven research
• Asking LLMs directly why they didn’t recommend your brand surfaces specific, actionable gaps that ranking data never could
• Enterprise audiences don’t share one AI tool the way marketers often assume—access varies by household, workplace, and region
• The fundamentals behind AI visibility trace back to classic marketing: product, price, place, and promotion, not just content and links
• Model consensus and disagreement across LLMs is itself a useful signal for knowing where to focus next

Two Decades Watching Search Evolve at One Company

Jason has stayed at SAP through mobile’s takeover of search, social media’s rise, and now the shift toward AI discovery—long enough to watch trends arrive with predictions of total disruption and settle into just one more channel to manage. What’s kept him there, he said, is less about any single trend and more about the relationships and institutional knowledge built up managing search at global scale, across a company present in nearly every market.

That scale gives SAP a rare vantage point. Feature rollouts that start in the U.S. often show up in SAP’s own data as they expand—AI Overviews launching in France, for instance, with enough volume behind them to measure real impact almost immediately.

On what makes this moment exciting rather than just exhausting, Jason didn’t point to a new tool. He pointed back to an undergraduate marketing class, where he mastered the four Ps:

• Product: direct feedback on how the market actually perceives an offering, not just whether it ranks
• Price: visibility into how AI systems describe and position pricing relative to competitors
• Place: clarity on which platforms and models are actually surfacing a brand to the audience that matters
• Promotion: the traditional home of SEO, now informed by the other three instead of operating in isolation

SEO, in his telling, has always lived almost entirely in promotion, and AI search is finally giving that work a way to feed insight back into the other three.

“We’re actually being able to send stuff back saying, hey, look, this is how the audience perceives the product,” Jason said. “That’s something we couldn’t do with just rankings alone.”

Splitting Time Between Technical and Content

One of Jason’s earliest decisions on SAP’s global SEO team was to deliberately divide attention between technical SEO and content, rather than letting easier-to-measure technical wins dominate the roadmap. That split became foundational well before AI search existed, and it’s held up as the requirements shifted.

The technical layer—hreflang tags, schema markup, page speed, crawlability—still determines whether AI systems can read a page accurately at all. 

“The more efficient the experience, and the easier it is for them to have a precise understanding of what’s on the page, the easier it is for them to make it easier for you to be successful,” he said.

From Reference to Inference

Jason described a shift he’s also been tracking closely: search moving from a reference model to an inference model. 

The reference era meant typing a keyword, getting a list of links, and doing the synthesis yourself. Inference flips that—people now describe a full situation and expect the system to reason through it.

He compared it to searching for a medical solution. “I don’t start searching on the medicine name. I start searching on the symptoms they have, to try and figure out what I need,” he said. 

Prompts built this way don’t map cleanly onto keyword research, since they’re longer, more situational, and carry context that used to be left unstated in a search box.

Why Ranking Number One Stopped Feeling Like Winning

Jason has watched keywords hold the top ranking position for years while barely getting clicked—a pattern familiar to anyone who’s worked in search long enough. What’s changed is that there’s now a faster, more direct way to find out whether that visibility actually matters.

“With LLMs, being able to ask them specifically, ‘ Why didn’t you recommend my solution or my brand?’ you’re suddenly realizing that rankings didn’t tell the full story,” he said. 

In other words, rankings haven’t become irrelevant; they’ve just stopped being the headline metric worth getting excited about on their own.

The One Question Every Brand Should Be Asking

That direct question—why didn’t you recommend us?—has become the closest thing Jason has found to a repeatable audit. His team identifies where SAP is and isn’t being cited for a given solution, then asks the models why, aggregating the answers across enough prompts to spot patterns.

The answers aren’t always accurate, but they’re consistently useful. Sometimes the fix is technical—content buried in a way that’s hard for a model to parse, or worth tagging more precisely with schema. Other times it’s a plain information gap: the answer exists, just trapped in a PDF or an untranscribed video. 

“I encourage everybody to do this,” Jason said. “You get really inspired really quickly about, okay, what can we do?”

The Household Test for Audience Fragmentation

One of Jason’s sharpest points came from his own home rather than a dashboard. Marketers tend to assume their own AI habits mirror their audience’s, and Jason’s household proves how wrong that assumption can be: his kids use YouTube as their search engine, his wife favors one LLM, and he uses another entirely.

“Just imagine what happens in the audience,” he said, “let alone that certain companies might be deploying their own AI models or restricting people’s access.” 

That fragmentation is exactly why he flags Microsoft Copilot as underweighted by B2B marketers—not because it’s dominant, but because plenty of enterprise buyers simply don’t have another option at work.

Watching Where the Models Agree, and Where They Don’t

With SAP’s product portfolio spanning dozens of solutions across a fragmented model landscape, Jason treats consensus itself as a signal worth tracking. When every model describes a product the same way, that’s a reasonable baseline. When they diverge, it usually points to exactly where to start digging.

“We just have to adapt and keep an eye on things,” he said. “Part of that is making sure that we’re tracking the right prompts over time”—consistent enough to actually reveal a trend, rather than a fresh set of prompts every time a new model ships.

Rankings Aren’t Dead. They’re Just Not the Headline Anymore

Looking years ahead, Jason didn’t predict the end of rankings, just their demotion. Positions will still be measured, but they’ll stop being the main story worth telling. 

“The main question will be, are you being recommended or not, are you being recognized,” he said. “A ranking was a recommendation about where to click through. A citation is a recommendation about where to validate the answer.”

Every shift Jason described in this conversation—reference giving way to inference, rankings losing their headline status, audiences splintering across models nobody assumed they’d use—traces back to the same four Ps he learned in a university marketing class. AI search didn’t replace that framework; it just gave it a faster feedback loop than rankings ever could.

The brands that keep showing up in AI-generated answers won’t be the ones chasing the newest model update. They’ll be the ones who keep asking why, and actually act on what comes back.

Voices of Search is a daily SEO and content marketing podcast hosted by Jordan Koene and Tyson Stockton. The show delivers actionable strategies and data-driven insights to help marketers navigate the ever-evolving world of search engine optimization and content marketing. New episodes air weekly, covering everything from technical SEO to AI discovery, featuring industry leaders and practitioners sharing real-world frameworks and proven tactics.

Subscribe to Voices of Search on Apple Podcasts, Spotify, or your favorite podcast platform. Follow Previsible on LinkedIn for updates and subscribe to the VOS YouTube channel for video episodes and clips.