Google AI Mode is citing Google. What brands should do next

Google AI Mode is not just citing websites. For a growing set of local and product queries, it is citing Google-hosted surfaces that represent your business instead of your own URL. According to Profound, google.com became the second-most-cited domain in AI Mode between April 15 and June 30, and its citation share grew 8.4x over that period. Most of that growth came from Business Profile cards and Product Knowledge Panels. For brands, the implication is blunt: AI visibility now depends partly on data layers you do not fully own, but can still shape.

That changes the playbook. If a restaurant, clinic, home service company, retailer, or manufacturer still treats its profile data and product feeds as secondary cleanup work, it is probably underinvesting in one of the inputs AI Mode uses most visibly. Your site still matters. But for certain high-intent searches, the cited asset may be Google’s version of your business, not your homepage or product page.

What changed in Google AI Mode citations?

The clearest change is that google.com moved up because AI Mode increasingly points to Google-hosted cards, not because Google suddenly started citing generic search results everywhere. Citation share measures how often a domain appears as a cited source in AI Mode answers. In Profound’s data, the jump came overwhelmingly from two subpaths: Business Profiles for local intent and Product Knowledge Panels for product intent.

That concentration matters. It suggests AI Mode is not redistributing citations evenly across all query classes. The shift is strongest when the answer can be assembled from structured facts such as opening hours, reviews, photos, product specs, or compatibility details. A local search like “best pediatric dentist near me open Saturday” and a product search like “compare air fryer basket sizes” are very different user needs, but both can be answered through a card-style surface faster than through a traditional page click.

This also fits a broader pattern. Google’s AI experiences have already shown a habit of citing Google-owned surfaces when they can satisfy the question directly. The new wrinkle is not that self-citation exists. It is that it is becoming more concentrated around data products that brands actively feed, update, and sometimes neglect.

Why are Google-hosted pages winning these citations?

Because they are structured, compressed, and ready for an AI system to reuse. A Business Profile is a tidy bundle of business facts. A Product Knowledge Panel is a tidy bundle of commercial facts. When AI Mode needs to answer a short, high-confidence question, those bundles are efficient.

Take a hotel query. If a user asks about location, hours, ratings, and photos, AI Mode can cite a Google-hosted business card that already packages those details. Take a product query. If a user asks whether two devices are compatible or which version has a specific feature, a Product Knowledge Panel can surface normalized attributes without making the user compare several retail pages.

That does not mean Google-hosted surfaces are always better than a brand’s own pages. It means they are often easier for the system to reference when the question is narrow and the answer is mostly factual. This approach works well for queries built on structured attributes. It is less complete for questions that need brand nuance, positioning, warranties, service differences, or category education. That limit matters, because it shows where your own content still has a real job to do.

What does this mean for local brands?

For local brands, Business Profile data is no longer a side file maintained by whoever has ten spare minutes on Friday. It is part of the AI search layer. Profound found the biggest citation changes in hospitality and travel, home services, restaurants and dining, real estate, and healthcare. Those are all categories where users ask fast, practical questions and expect AI Mode to answer without friction.

Imagine a multi-location urgent care group. A user asks which clinic is open late, accepts walk-ins, and has strong reviews. If AI Mode cites a Business Profile card instead of the clinic’s location page, the brand is still present, but the source of truth in the answer has shifted. The same goes for a restaurant chain when users ask about hours, menu photos, or nearby options. If the profile is outdated or thin, the brand can appear less useful even when its website is excellent.

That makes local hygiene a visibility issue, not just a maps issue. At minimum, teams should review:

• opening hours and special hours
• photos that reflect the current location experience
• review quality and recurring review themes
• consistency between the website’s local pages and the profile facts users see in Google

The last point is easy to miss. AI systems do not love ambiguity. If your site says one thing and the profile suggests another, the model has to reconcile the conflict. That is rarely good for recommendation confidence. Brands that want stronger local AI visibility should read how local AI search now treats your website as the source of truth because the best results come when the site and the profile reinforce each other, not when one silently contradicts the other.

What does it mean for ecommerce and product teams?

For ecommerce teams, the message is similar but the input layer is different. Product Knowledge Panels are gaining citation share for comparison, compatibility, and specification queries. In plain English, AI Mode may cite a Google-hosted product surface when the user asks the kind of question that used to send them into a maze of retailer tabs.

Consider a shopper asking whether a charger works with a certain phone model, or whether one vacuum is lighter than another. Those questions rely on structured product facts. If your feed data is incomplete, inconsistent, or slow to update, you are giving AI Mode weaker material to reference. A beautiful product detail page still matters for persuasion, conversion, and deeper education, but it may not be the cited surface for these narrow queries.

This is why feed quality deserves a bigger seat at the visibility table. Titles, attributes, specifications, availability signals, and variant clarity are no longer just shopping operations details. They influence whether your product can be represented cleanly inside an AI answer. Teams that want a sharper view of this shift should also look at Google’s Merchant Center becoming an AI shopping visibility dashboard, because the operational boundary between shopping data and AI discoverability is getting thinner.

There is also an important nuance here. Profound interprets the rise of Product Knowledge Panel citations as an early sign of Google’s Universal Commerce Protocol. Google describes UCP as an open standard for direct purchases across AI Mode and Gemini, with retailers remaining the merchant of record. But the connection between the cited panels in this dataset and a UCP-enabled buying flow is still an interpretation, not official confirmation. Smart teams will pay attention without pretending the evidence is stronger than it is.

BotRank’s Take

This story is a good example of why AI visibility cannot be measured with a rank tracker alone. When a brand asks, “Did AI Mode cite us?” the honest follow-up question is, “Which version of us?” Sometimes it is your website. Sometimes it is a marketplace listing. Sometimes it is a review site. And increasingly, for local and product intent, it can be a Google-hosted surface that represents your business without sending citation credit to your domain.

That is where BotRank’s Source Analysis feature becomes genuinely useful. It does not just count mentions. It shows which domains and pages are actually being cited, how that differs by model, and whether the cited page really mentions your brand in a useful way. In a case like this, that distinction matters. If AI Mode keeps citing a Google surface while ChatGPT or Perplexity cites your own page, the optimization brief is not “write more content.” It is “understand which source layer each engine trusts, then fix the weakest one first.”

How should SEO and GEO teams respond now?

The practical response is not to panic about lost clicks. It is to expand the unit of optimization. GEO, or generative engine optimization, is the practice of improving how your brand appears in AI-generated answers. In this case, GEO means managing the data sources AI Mode is most likely to cite for local and product questions, while still maintaining strong first-party pages.

A solid response plan usually includes five moves:

• Separate your prompt sets by intent. Track local, comparison, compatibility, specification, and branded discovery queries separately. These behaviors do not pull from the same source layers.
• Audit the inputs behind the card. For local brands, that means Business Profile completeness and accuracy. For product teams, it means feed quality and structured product attributes.
• Compare citations to owned pages. If AI Mode cites a Google-hosted surface but other engines cite your site, you need to know that before you decide where to invest.
• Fix contradictions between systems. Location pages, support docs, and product pages should agree with the facts exposed in profiles and feeds.
• Measure over time, not once. AI citation patterns can shift quickly, especially when a platform changes how it formats local and commercial answers.

This is exactly why AI Visibility tracking across multiple LLMs matters. A brand should not assume Google AI Mode behaves like ChatGPT, Gemini, or Perplexity on the same prompt. One model may reward a clean local page. Another may lean on a platform profile. A third may cite editorial reviews instead. If you only watch one surface, you can easily optimize the wrong thing.

There is also a workflow angle here. Once you see the weak spots, you need a way to turn them into prioritized fixes. That is where BotRank’s technical audits become practical. They help teams translate an abstract visibility problem into concrete actions, whether the issue is a thin product page, a crawl blocker, or a mismatch between structured data and what AI systems are surfacing.

If you want a broader framework for this shift, AI visibility starts before the prompt and ends with citations is the right mindset. The user only sees the final answer. Your job is to manage the full evidence trail that leads to that answer.

What this does and does not mean

It does mean that AI search visibility is getting more platform-mediated. For many local and product queries, Google is increasingly comfortable citing its own hosted representations of businesses and products. That puts pressure on publishers and brands because the citation can land on Google’s surface instead of a first-party page.

It does not automatically mean your website traffic was replaced one-for-one. Profound’s citation-share data describes what AI Mode links to, not where the user ultimately clicks, converts, or buys. A cited Business Profile card may still help a local brand. A cited product panel may still help a retailer. But those wins are different from a traditional page visit, and they need different measurement.

It also does not mean websites matter less across the board. For local trust, your site is still the place where AI systems can validate brand claims, service details, and deeper expertise. For ecommerce, your product pages still do the heavy lifting on education, differentiation, proof, and conversion. The smarter interpretation is narrower: for some query classes, structured surfaces are becoming the citation layer, while first-party pages remain the explanation and conversion layer.

That split is uncomfortable for teams raised on URL rankings, but it is easier to work with once you accept it. A restaurant group does not win by choosing between its website and its profile. It wins when both say the same thing clearly. A retailer does not win by choosing between product feeds and product pages. It wins when both are complete enough that AI systems can trust the facts and shoppers can trust the brand.

If you are seeing strong SEO performance but messy AI attribution, you are not alone. BotRank has already written about why high Google traffic does not equal AI citations. This is the same problem wearing a more specific outfit: visibility can shift to a different surface even when the brand itself is still relevant.

FAQ: what brands should know about Google-hosted AI citations

Does a google.com citation mean I lost the click?

Not necessarily. Citation share shows what AI Mode links to in the answer, not where the user ultimately goes next or whether they convert.

Should local brands focus on Google Business Profile or their website first?

They need both, but for different jobs. The profile often supports quick factual answers, while the website remains the deeper trust and conversion layer.

Do product feeds matter more than product pages now?

For comparison, compatibility, and specification questions, feed quality may have a larger influence on the cited surface. Product pages still matter for proof, nuance, and sales.

Is this only a Google problem?

No. Every AI system has its own citation habits. The Google-specific twist is that AI Mode can lean heavily on Google-hosted surfaces for local and commercial intent.

How can I measure this without guessing?

Track the same prompt set across multiple models and inspect the cited pages, not just the brand mentions. If you can see whether the citation lands on your site, a Google card, or a third-party page, you can prioritize the right fix.

The takeaway is simple. In AI search, the winning asset is not always the page you wanted cited. For local and product intent, the smarter play is to optimize the data objects Google can surface, keep them aligned with your first-party pages, and measure how each model turns that evidence into visibility. If you want to see that source mix clearly, BotRank is built for exactly that job.