AI Mode at 1B, Brands Invisible & Conversational Ads

Search Digest Marc Levy Week of 25 May 2026 Issue #012

The Google AI Mode usage data and the Victorious brand visibility study sit next to each other in a way that tells one story. AI Mode has over a billion monthly users. Queries have tripled in length. Planning intent is surging. The channel is genuinely large and growing fast. And 90% of brands are completely absent from it. The gap between the scale of the audience and the number of brands earning any presence in AI answers is wider than most teams have assumed.

The Lily Ray data on AI content that backfires is the context for why that gap does not close quickly. 54% of sites that scaled AI content experienced 30% or more organic traffic declines, often within a year of their content peak. The Kevin Indig piece on Microsoft's index shift makes the same point from a different angle: the index is being rebuilt around groundable, attributable facts, not document volume. Scaling into AI visibility without building the right kind of specific, attributed content produces the same pattern Ray describes.

The Google conversational ad formats announcement shows where the commercial pressure is coming from. Highlighted Answers embed a sponsored result inside the same recommendation list that organic citations occupy. Conversational Discovery ads appear directly inside AI Mode answers. As those formats scale, the competitive space for organic content in commercial queries gets tighter. Building AI visibility now, before the ad formats dominate the answer environment, is a better position to be in than building it after.

Google / blog.google

How AI Mode Is Changing and Expanding the Way People Search

The official first-party numbers arrived on 19 May as part of the Google I/O announcements. AI Mode has over 1 billion monthly active users globally, with queries more than doubling every quarter since launch. The query length data stands out: the average AI Mode search is three times longer than a traditional Google query. More than one in six searches in the US now uses voice or image input, and image searches grew 40% month-on-month. These are not third-party estimates. They are Google's own measurements shared publicly at I/O for the first time.

The planning query data shows something useful about changing intent patterns. Queries related to planning grew 80% faster than overall AI Mode usage in the past six months. Brainstorming queries grew 30% faster. What this points to is that people are bringing longer, more exploratory, more multi-step questions to AI Mode than they ever brought to traditional search. A page that answers a single keyword phrase and stops will satisfy a traditional search result. It will not serve someone asking a planning question in AI Mode, where the follow-up questions are part of the same session. Content built to support a full intent journey, not just a single query, is better positioned as AI Mode grows.

Key points

AI Mode has over 1 billion monthly active users globally after one year, with queries doubling each quarter since launch
Average AI Mode query is three times longer than a traditional Google search
More than 16% of US searches now use multimodal inputs: voice, images, or files
Image searches in AI Mode grew 40% month-on-month
Planning queries grew 80% faster than overall AI Mode queries in the past six months
Brainstorming queries grew 30% faster than overall queries; Google reports search volume is at an all-time high

Key takeaway

Content structured to answer extended, exploratory questions is better placed as AI Mode grows. If your pages answer a single question and stop, they may satisfy a traditional search intent but miss the follow-up questions AI Mode users ask in the same session. Build content that supports a planning journey rather than a standalone query, and make sure the most important claims appear early, since AI Mode users are asking questions that require the answer quickly rather than reading to a conclusion.

Also worth considering

The query length increase has implications for keyword research. Keywords drawn from traditional Search Console data, where queries average two to three words, will not reflect how users phrase questions in AI Mode. If you are building a prompt tracking strategy alongside traditional keyword research, the prompts should reflect the more conversational, intent-driven format that AI Mode users actually use, not a compressed keyword version of the same question.

What I'm testing

Looking at whether Search Console data shows a difference in average query length for pages that appear in AI Mode versus pages that do not. If AI Mode is drawing longer queries as Google reports, there should be some evidence of it in non-branded query data over recent months. The specific question is whether content structured around intent journeys versus single-query answers shows a different query length distribution.

Read the full article

Michael Transon / Search Engine Journal (Victorious research)

90% Of Brands Have Zero AI Search Mentions, New Study Finds 4 Key SEO Insights

Victorious published research on Search Engine Journal this week covering brand visibility across eight AI platforms in Q1 2026. The headline number: 89.8% of the 177 brands tested registered zero mentions across any of the eight platforms. Only 18 brands had any mention rate above zero. The study tested ChatGPT, Perplexity, Gemini, Google AI Overviews, Microsoft Copilot, Claude, Meta AI, and one further platform across 107,011 total AI responses. Brands came from five verticals: healthcare, SaaS, financial services, ecommerce, and legal services.

The vertical patterns are as useful as the headline figure. Healthcare, SaaS, and financial services brands showed consistent mentions and citations. Ecommerce brands got more mentions than citations, meaning they appeared in answers but were less often linked. Legal services showed the reverse: citations without brand attribution. That distinction matters for how you measure AI visibility. A citation without a brand mention drives traffic but does not build the name recognition that leads to repeat visits. The study also found that the competitive gaps in most verticals are still wide open. With only 18 brands earning any AI visibility across the entire dataset, there are very few incumbents to displace if you decide to build for this channel.

Key points

89.8% of 177 brands tested had zero AI mentions across 8 platforms in Q1 2026
Only 18 brands registered any AI mention rate above zero, across 107,011 total AI responses
Eight platforms tested: ChatGPT, Perplexity, Gemini, Google AI Overviews, Copilot, Claude, Meta AI, and one more
Healthcare, SaaS, and financial services showed consistent both mentions and citations
Ecommerce: more mentions than citations; legal services: citations without brand attribution
Most verticals have very few incumbents holding AI visibility, meaning competitive gaps have not yet closed

Key takeaway

Before you can improve AI visibility, you need to know where you stand across the platforms your audience uses. Performance on Google AI Overviews says nothing about visibility in ChatGPT, and the two are independent signals. Running a baseline measurement across your most important platforms, using varied prompts as the SparkToro research from last week recommended, is the starting point for any strategic response. The competitive gap is still wide in most verticals, so the measurement effort is worth it.

Also worth considering

The ecommerce and legal services distinction, where one vertical gets mentions without links and the other gets links without brand names, points to something deeper in how AI platforms handle commercial queries differently from professional ones. Understanding which pattern applies to your sector changes where you focus your optimisation. If you are getting citations without brand attribution, entity clarity and brand name consistency across your content are the levers to pull first.

What I'm testing

Running a visibility check across five platforms using identical brand queries, to see whether the vertical patterns in this study hold at smaller scale. The research used 177 brands so the averages hide a lot of variation at the individual level. I want to understand whether the absence rate is consistent across query types or concentrated in specific prompt formats, since that would change the content strategy for improving visibility.

Read the full article

Lily Ray / Substack

It Works Until It Doesn't: AI Content Strategies That Backfire

Lily Ray tracked 220-plus domains from the publicly available case study pages of AI content platforms, looking at what happened to their organic traffic over time. The pattern she found is consistent across the dataset. Rapid AI content growth, typically over 6 to 12 months, is followed by a traffic peak. That peak arrives 3 to 6 months after the content peak. Then a steep decline, often falling below the baseline level the site started from. 54% of the tracked sites experienced 30% or more organic traffic declines. 39% lost 50% or more of their peak traffic. 22% lost more than 75%.

The eight patterns she identifies include comparison pages at scale, glossary factories, programmatic location and language pages, FAQ farms, competitor-versus-alternatives pages, and off-topic content built for keyword volume. What changes in 2025-2026 is the speed of the cycle. AI tools make it faster to produce content at volume, so sites accumulate the signals Google identifies as scaled, low-quality content faster. The decline follows, often within a year of the peak. Ray is careful to note these are third-party traffic estimates, not verified site analytics, and the data reflects correlation rather than confirmed causation. But the pattern is consistent enough across 220-plus domains to be instructive for anyone building an AI content programme.

Key points

54% of 220-plus tracked sites experienced 30%+ organic traffic declines after scaling AI content
39% lost more than 50% of peak traffic; 22% lost more than 75%
Typical trajectory: content growth for 6-12 months, traffic peak 3-6 months after content peak, then steep decline
Eight patterns identified: comparison pages at scale, glossary factories, programmatic location/language pages, FAQ farms, self-promotional listicles, competitor-vs-alternatives pages, off-topic content, and brainstorming pages
Declines often take traffic below the site's pre-AI-content baseline, not just to a lower growth level
Data sources: Ahrefs organic traffic estimates and Sistrix Visibility Index across AI platform case study domains

Key takeaway

The risk with AI content scaling is not that Google will eventually penalise it. The risk, based on Ray's data, is that the drop happens quickly and typically takes sites below where they started. For any content programme using AI at scale, the relevant question is not whether it is working now, but whether the traffic is coming from genuine audience demand or from patterns Google's quality systems are learning to identify. Content built to answer specific audience questions is more durable than content built to capture keyword volume.

Also worth considering

Ray's eight patterns are not about AI authorship specifically. They are patterns that produce thin, low-value content regardless of how it is written. AI makes them faster to execute, which is why the cycles are shorter and the drops are steeper. That framing is useful in team conversations about AI content strategy. The risk is not using AI to write content. It is publishing at volume without a quality filter that reflects genuine audience need.

What I'm testing

Checking whether any content types in my tracking data match the patterns Ray identifies, specifically looking at whether comparison or glossary content added in the last 12 months shows early signs of the trajectory she describes. The leading indicator I am watching is the timing between content volume peak and traffic peak. If the traffic peak arrived well after the content peak, that fits the pattern and is a signal to watch the trend carefully over the next six months.

Read the full article

Anu Adegbola / Search Engine Land

Google Tests New Conversational Ad Formats in AI Mode and Search

Google announced four new ad formats for AI Mode and Search at Marketing Live 2026, all powered by Gemini. Conversational Discovery ads answer specific queries directly in AI Mode. Highlighted Answers appear within AI-generated recommendation lists, placing a sponsored result inside the same format organic citations use. AI-powered Shopping ads generate dynamic product explainers for high-consideration purchases rather than serving a static ad. Business Agent for Leads replaces the contact form with a Gemini-powered chat experience that qualifies leads and captures information in conversation. Two of these formats are in US testing now. Shopping ads and the Business Agent follow later in 2026.

The architecture is significant. Conversational Discovery and Highlighted Answers formats sit inside the AI-generated answer, not alongside it. That changes the relationship between ad placement and user intent in a way that keyword-targeted text ads do not capture. In traditional search, an ad competes with organic results for attention. In these formats, the ad is embedded within the answer itself. For advertisers, the implication is that creative needs to work within the conversational context of a query, not as a standalone call to action. Gemini generates ad creative dynamically based on the specific conversation, so the same brand asset can produce different outputs depending on how the question is phrased.

Key points

Four new formats announced: Conversational Discovery ads, Highlighted Answers, AI-powered Shopping ads, and Business Agent for Leads
Conversational Discovery and Highlighted Answers are in US testing on mobile and desktop now
Highlighted Answers embed the sponsored result within the AI-generated recommendation list, not alongside it
Gemini generates ad creative dynamically based on the specific user query rather than serving pre-written copy
Business Agent replaces static contact forms with a Gemini-powered chat that qualifies leads in conversation
AI-powered Shopping ads generate product-specific explainers for high-consideration purchases; rolling out later in 2026

Key takeaway

The Highlighted Answers format is the one to watch most closely, because it changes what organic means in AI search results. If a sponsored result appears within the same recommendation list AI generates from organic content, the visual distinction for users narrows. Understanding how Highlighted Answers perform relative to organic citations, once data is available, will be important for understanding the true competitive effect on organic visibility in commercial query spaces.

Also worth considering

Business Agent for Leads is a meaningful commercial format for professional services. A Gemini-powered chat that qualifies leads before they reach a human is a real conversion tool, not just an awareness format. For sectors where the buying decision involves a discovery conversation, this has direct revenue implications. It also shifts where the qualification conversation happens: from your website into Google's AI environment. That is a meaningful change for how conversion paths work.

What I'm testing

Watching for Highlighted Answers to appear in SERPs for commercial queries in the niches I track. The specific question is how often they appear in answer formats that currently show only organic citations, and whether the visual treatment is distinct enough for users to recognise the placement as advertising. The labelling and user perception angle is worth following, since those questions will shape how this format competes with organic over time.

Read the full article

Kevin Indig / Growth Memo

Growth Intelligence Brief #18

Microsoft published a statement on 6 May defining what the search index is now for. The framing Kevin Indig covers in Growth Intelligence Brief #18 is straightforward and significant: the index goal is no longer "find the best documents." It is "find the best information to support synthesised answers." Indig identifies three resulting shifts. The primary quality metric changes from relevance to factual accuracy. The goal changes from document retrieval to extraction of what Microsoft calls "groundable information," discrete, attributable facts that AI systems can cite responsibly. Content freshness carries more weight because AI systems prioritise current, accurate facts over evergreen documents.

The practical direction is to audit content for three things: whether it contains distinct, attributable claims; whether author, date, and source are clearly visible; and whether AI crawler access is unrestricted. Indig recommends that product and category pages should front-load key facts, placing the most important claims in clean text within the first 600 words. That is the range where AI parsers are most likely to retrieve content for inclusion in synthesised answers. The three-step optimisation model, crawl, parse, retrieve, reframes where to focus effort. Getting crawled is table stakes. Getting parsed accurately is the bottleneck for most sites. Getting retrieved into an answer is the outcome that drives AI mentions.

Key points

Microsoft officially defined its search index goal as "supporting answers" rather than "ranking pages" on 6 May 2026
Primary quality metric shifts from relevance to factual accuracy and attribution
"Groundable information" is the target: discrete, attributable facts an AI system can cite accurately
Content freshness carries more weight in AI-grounded answers than in traditional document ranking
Key facts should appear in clean text within the first 600 words for reliable AI parser retrieval
Three-step AI optimisation model: crawl (table stakes), parse (main bottleneck), retrieve (the outcome)

Key takeaway

Check whether your highest-priority product and category pages have their most important claims in the first 500 to 600 words, in clean text rather than in tables or behind expand components. That is the most direct change this piece points to. It does not require a content rewrite, just restructuring where key facts appear. For most sites, the important claims are buried in a third section or framed too generally to be retrieved as groundable information. Moving them up is a low-effort change with a clear purpose.

Also worth considering

The shift to groundable information changes what good content looks like at the claim level. "Our platform reduces churn" is not groundable. "Companies using our platform report a 32% average reduction in churn over 12 months" is. The distinction between vague marketing language and specific, attributable, verifiable claims maps exactly onto what AI systems are being built to retrieve. Brand and content teams writing for AI audiences need to understand that distinction and apply it to every key claim on their most important pages.

What I'm testing

Auditing a small set of pages to check whether the key claims are specific enough to be groundable, meaning they include a number, a timeframe, or an attributed source, and whether those claims appear within the first 600 words. Most pages I look at have important claims buried in a later section or framed too generally. Reordering that content to front-load the attributable facts is a straightforward change, and I want to track whether it affects AI visibility over a 60 to 90-day window.

Read the full article

The core tension in this week's reads is between scale and quality. The AI Mode usage data shows a channel growing faster than most teams are resourced to address. The brands study shows how few have got any traction. Lily Ray's data shows what happens when you try to close that gap with content volume. The Microsoft search index framing from Kevin Indig gives the clearest operational answer: factual accuracy, attribution, and provenance are now quality signals in their own right. Build for groundable information and you are building for the index as it actually works in 2026.

The conversational ad formats are a signal about where things go from here. Highlighted Answers placing a sponsored result inside the organic answer list is not a small change. It is a preview of the competitive environment organic content will be operating in once these formats scale. The brands that have built genuine AI visibility now will have the citation history and entity clarity to hold ground in that environment. That is harder to acquire with ad spend than it is to build with content.

Free Consultation

Let's Talk

Tell me what you're working on. I'll give you an honest assessment and we'll explore if working together makes sense — no hard sell, just a free, no-obligation call to explore what's possible for your business.