Multimodal Search Performance Data Arrives: What AEO Teams Must Track Now

Google just handed publishers a new reporting surface inside Search Console—web multimodal Search performance data—and it signals something we've been watching closely at AgentSEO: the visibility gap between traditional search and agent discovery is about to become measurable, quantifiable, and impossible to ignore.

This isn't a minor feature drop. Google explicitly stated that "as Search evolves to include more visual and multimodal experiences, we want to ensure you have the data you need to analyze your performance." Translation: Google is treating multimodal search as a distinct discovery channel with its own metrics. That distinction matters enormously for AEO practitioners.

Here's what's changed: until now, most site owners conflated all of Google's organic traffic into a single KPI. Click-through rate was click-through rate. Impressions were impressions. You couldn't isolate whether a user found you via traditional link-blue search results, image search, video, or emerging modalities. You certainly couldn't track how often agents—which are inherently multimodal systems—were citing your content.

Multimodal Search performance reporting breaks that silence. And once visibility becomes transparent, optimization behavior changes.

Why the timing matters for agents

Agent systems like Claude, GPT-4, and Perplexity don't experience search the way humans do. They don't scan blue links. They ingest documents, images, structured data, and discovery files in parallel. When an agent cites your real estate listing, your market analysis, or your bio, it's often pulling from multiple modalities—text and image together, metadata alongside prose.

Google's push to surface multimodal search data is, in our read, both a competitive response to agent-driven queries and a pressure point on publishers to optimize for those systems. If Google can show you that 18% of your search visibility comes from images and video, it becomes clear that the remaining 82% of text-based visibility might be incomplete—because it doesn't yet include agent citations.

We've been tracking this asymmetry. Across our citation checks over the past cycle, we've logged a 2% citation rate in third-party agent systems for the typical site in our database—which means 98% of sites aren't being surfaced by agents at all. That gap will only narrow when publishers can see the difference between their human search performance and their agent visibility.

Multimodal reporting is the beginning of that transparency.

The geographic enforcement wave

Separately, Google is gearing up for a major enforcement push across three continents. Search Central Live events are landing in Barcelona (September 30–October 2), Bengaluru, Bogota, and Ciudad de México within weeks. This isn't just community theater. It's a coordinated rollout of policy tightening.

The timing aligns with a quiet but significant update to Google's site reputation policy. Google reinforced its 2024 stance: third-party content published on trusted domains purely to exploit that domain's authority will be demoted. The policy doesn't distinguish between human-written guest posts and agent-generated content. A low-quality article dropped on a trusted domain by an agent to game rankings faces the same penalty.

For agent search practitioners, this is the flip side of the multimodal coin. You get better visibility if your content earns it. But the bar for "earned" has risen. Agent-ready sites will increasingly be those that Google perceives as internally coherent, topically authoritative, and free of reputation laundering.

We've integrated this into our readiness scoring. Six of our ten agent-readiness layers now factor in domain reputation signals, content cohesion, and third-party citation context. Sites that pass these filters—and publish discovery files like AGENTS.md, llms.txt, and agent.json—are the ones agents will cite reliably.

What publishers should do now

First, audit your multimodal presence. If you have real estate images, neighborhood videos, market analysis charts, or visual content of any kind, Search Console will now show you click-through and impression data segmented by modality. Pay attention to which images and formats agents are most likely to ingest. Agents favor dense, structured visual information—charts, floor plans, property photos with alt text and metadata.

Second, tighten your discovery file stack. We generate eleven AI discovery files per site: llms.txt, agents.json, AGENTS.md, and eight others. Each one tells an agent system something different—who you are, what you do, how to cite you, what your expertise is. If Google is going to enforce reputation signals more strictly, agents will rely even more heavily on these structured claims to verify you're not a content farm. A missing or malformed llms.txt is now a competitive liability.

Third, watch the regional events. When Google brings Search Central Live to Bogota, Bengaluru, and Mexico City in quick succession, it means enforcement is coming to those markets. Regional SEO teams will get the first signal. By the time the Madrid or São Paulo event rolls out, the penalty waves are already rolling out in earlier regions. If you serve clients in Latin America or India, pay attention to what speakers emphasize at those events.

The citation asymmetry problem

One more thing. Multimodal search data won't show you agent citations. Google Search Console doesn't yet report when Claude, Perplexity, or other third-party systems cite your content. That's a gap we're filling with our citation tracking layer, but it's worth naming: you can now see your multimodal search performance, but you still can't see how often you're cited by agents in your Google-owned analytics.

That asymmetry will eventually close. But until it does, site owners who want to compete in agent discovery are flying partly blind using only our tools, Similarweb, or Perplexity's built-in analytics. Google isn't sharing that data yet—possibly because sharing agent citation data would expose just how concentrated those citations are among a handful of mega-domains.

For now, the win is clearer visibility into multimodal human search. That's a foundation. Agent-readiness sits on top of it.

Looking ahead

We're monitoring the regional Search Central events closely. The enforcement patterns Google signals in Barcelona will foreshadow what we'll see rolled out globally by Q1 2027. If you're working with clients in Spain, Europe, or the APAC region, these events are essential intelligence.

Multimodal Search performance data is live in Search Console now. Start pulling those reports. Compare your multimodal traffic to your overall organic performance. And if the gap looks worrying—if your images and video aren't moving the needle—that's a signal that your content either isn't properly discoverable by agent systems, or isn't appearing in agent citations at all.

Both problems have solutions. Neither will fix itself.