New research from Adobe-owned brand visibility firm Semrush argues that AI visibility is shifting from individual prompts to broader topical authority.
Among the study’s findings:
- “Winning one prompt in ChatGPT is not the same as owning a topic. Real topic ownership requires a brand to appear across at least four of five related prompts, with a 5-percentage-point lead over the runner-up.”
- “Most topics in ChatGPT are still contestable.” Only 15.2% of the 1,094 categories analyzed had a clear owner, leaving 85% open to competitors.
- “Broad SEO signals do not consistently predict topic ownership.” Domain-level metrics such as Authority Score and organic traffic correlated with ownership only about half the time.
- “Strong SEO helps you show up, but… deep coverage of the specific topic is what turns you from runner-up to owner.”
- “Once you become the brand ChatGPT associates with a category, there’s a good chance you can hold it.” The study found category owners retained first place in 90.4% of month-over-month comparisons.
- “Your topical authority determines whether ChatGPT will surface you when a customer asks something new in that category.”
Read: “AI visibility is a topic-level game: A study of 50,000 brands in ChatGPT” (July 20) – Semrush blog
From tipsheet: AI appears to be changing the unit of competition from the individual webpage to the knowledge base. Rather than rewarding isolated articles, AI rewards publishers that build sustained expertise around a narrowly defined subject. The strategic asset is no longer the page. It’s the corpus. For example, the tipsheet corpus.
LLMs & CHATBOTS
AI spending’s accountability phase
Advertisers are beginning to scrutinize the cost of the compute behind AI as they embed it into daily workflows. Digiday reports that agencies and brands are introducing token caps, governance policies and new pricing models as AI usage moves from experimentation to everyday operations.
Digiday describes a conversation last month in Cannes:
“Put to a holding company executive, the same discomfort came out sharper. ‘Very few people are actually talking about the fact that the infrastructure has a real cost to it,’ they said. ‘Tokens have a cost, compute has a cost — if you just go wild, the cost of the machines will quickly outpace the cost of humans.’”
Read: “‘We’re starting to wonder’: Ad industry chases AI value as usage outpaces proof” (July 21) – Digiday (subscription)
From tipsheet: AI has created a capital allocation challenge that likely falls to the CFO. Compute is becoming another form of productive capital alongside labor and software. The organizations that win won’t necessarily spend less on AI, but they’ll generate more revenue and profit from their investment in AI infrastructure.
At the same time, an unwillingness to invest in AI could also signal an unwillingness to learn, experiment and discover the next competitive advantage.
SELL-SIDE
Standardizing AI’s programmatic infrastructure
On LinkedIn, Amazon Ads Director, Omnichannel Supply, Chris Conetta highlighted new capabilities in Amazon’s Certified Supply Exchange (CSE) program, which Amazon first announced in June.
Conetta highlighted yesterday:
- “What’s new? Five leading SSPs — Magnite, PubMatic, Index Exchange, OpenX, and Microsoft Monetize — are now certified across 8 technical areas including traffic optimization, signal quality, and automated deal activation.”
- “Two new innovations: Dynamic Traffic Engine (DTE) shares demand signals before bid requests are sent to surface more relevant inventory, and Live Events Optimizer (LEO) brings smarter budget allocation to live sports.”
Read more. (July 21)
The program builds on Amazon’s January introduction of Amazon DSP’s supply evaluation capabilities. As Conetta explained in June: “That principle doesn’t end at the demand-side platform (DSP) decision-making layer; rather, it extends to our proactive work with supply-side platforms (SSPs) to optimise the decision landscape.”
More: Amazon Ads Certified Supply Exchange programme advances the standard for open internet supply (June 21) – Amazon Ads
From tipsheet: By defining technical standards for SSP participation, signal quality and automated deal activation, Amazon is making more of the programmatic supply chain machine-readable and more useful for AI.
Related: Red Seat Ventures Announces Introduction of Premium Creator and Podcast Communities to Amazon DSP (July 20) – Fox Corporation
TECH
OpenAI ad revenue narrative
The prediction by eMarketer analyst Nate Elliott regarding OpenAI ad revenues through 2030 continues to reverberate across the media landscape.
Stories this week include:
- OpenAI faces stark doubts about ChatGPT ads. It’s beginning the long fight to prove them wrong. (July 21) – Business Insider
- OpenAI Appears to Be Missing Its Sales Goals by a Vast Margin (July 20) – Futurism
- OpenAI’s ad strategy faces a major reality check (July 21) – Fast Company (subscription)
Elliott believes OpenAI will perform far below its current projections. He wrote: “OpenAI thinks ChatGPT will collect about $50bn in US ad revenues in 2030. We think their entire addressable market will be 1/10th that.” Read his research launch post on LinkedIn. (June 4)
From tipsheet: The widespread attention to eMarketer’s forecast highlights how uncertain the future of AI advertising remains. Elliott’s model joins New Street analyst Dan Salmon’s projections as one of the few public attempts to quantify ChatGPT’s long-term advertising opportunity. Together, they are helping to frame the debate over OpenAI’s advertising ambitions, at least until additional forecasts emerge.
The eventual public release of OpenAI’s S-1 filing is likely to be the company’s most comprehensive opportunity to present its own case.
LLMs & CHATBOTS
Developments
- Microsoft and Mistral expand strategic partnership to give enterprises and regulated industries frontier AI they can control (July 21) – Microsoft
- Introducing Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber (July 21) – Google
- Meta testing StoryKit, an AI iPhone app that creates ‘personalized children’s stories’ (July 21) – 9to5Mac
AGENCIES
AI becomes part of the team
Ad Age Senior Technology & AI Editor Garrett Sloane highlighted how independent agency broadhead. is operationalizing AI to increase productivity without adding staff.
Sloane teased his article on LinkedIn:
“I spoke with broadhead.’s Maija Hoehn to get the lowdown on how indie shops can make technological changes in their organizations. We came away with a blueprint, about how to increase clients served by 30% on a smaller staff.
- ‘Everyone can have AI now. It’s the agencies that know how to use it and operationalize it,’ Maija told me.
- They got the team on board by launching an AI platform internally, and treating it like a product. It’s called Butter.
- Broadhead. used AI to cut down on hourly billed tasks like status reports.
- Budget for AI like it is a staff member, part of the labor costs.
- Then change the business model, once it’s all implemented.“
Read more. (July 21)
More:
- “A small agency’s 12-step guide to integrating AI across the business” (July 21) – Ad Age (subscription)
- “Everything Is Better with Butter” (February 2025) – broadhead.
From tipsheet: What could be some of the changes to the business model down the road? How about:
- Pricing model. If AI reduces labor hours, billing by the hour becomes less attractive. Agencies may shift toward retainers, value-based pricing or outcome-based pricing.
- Capacity model. Serve 30% more clients with the same staff, improving margins without raising fees.
- Service mix. Sell AI-enabled products, platforms or subscriptions instead of only custom services.
- Staffing model. Hire fewer junior employees and rely more on senior talent augmented by AI.
SELL-SIDE
Spotted in the wild: Google gen AI controls
Laurie Sullivan reported in MediaPost yesterday:
“Publishers have been looking for a way to keep their web pages out of AI Overviews, AI Mode, and generative AI features in Google Discover.
Now Google is testing a way that allows publishers to keep their content in traditional AI search results, as measurement companies develop new attribution models for AI-driven search.
A Google Search Console document describes how it began testing this option in June 2026. The control is now rolling out to select website owners.”
Read:
-
Google Gives Publishers Opt-Out For AI (July 21) – MediaPost
-
Search generative AI control – Google Search Console Help
More: Google separates AI Search from Search (June 4) – tipsheet
From tipsheet: In June, the UK Competition and Markets Authority (CMA) required Google to let publishers participate in Search while opting out of generative AI features, effectively establishing separate participation controls for generative AI uses of publisher content. The new controls demonstrate how regulation is beginning to shape the separation between AI Search and traditional Search.
MORE
- “Should Hotels Abandon Google Search Because of Zero-Click, Wholesalers Aren’t Worried About AI, APAC Hotel Investment Up 54%” – See community discussion on industry trade Hospitality Net
- “The Profound MCP is now available as an official Anthropic connector.” (July 21) – James Cadwallader, CEO, Profound on LinkedIn
- Opinion: “Beyond Keywords: How Agentic Commerce Search Understands True Shopper Intent” (July 21) – John Andrews on Salesforce 360 blog
- Opinion: Anthropic’s marketing success holds key lessons for B2B brands (July 14) – Jenny Sagstrom, CEO, Sköna on The Drum
- 2026 Mid-Year Commerce Media & Summer Sales Benchmark Report (July 21) – Pacvue (email required)