Agent Beat for September 15, 2026

Welcome to Agent Beat. It's Tuesday, September 15, 2026. This is your daily briefing on how AI finds brands and how agents choose what to buy. We follow the places where AI answers, ads, and shopping decisions meet. Today's stories all touch one question: when an AI system recommends or sells something, where does the money go, and how should brands plan for it?

A column in Stratechery says ChatGPT ads are working. The same column says Amazon ads placed in ChatGPT solve Amazon's biggest problem with chatbots. The column also covers Walmart's decision to accept Apple Pay. For marketers, the column's advice is direct. Treat ad placements inside ChatGPT, including Amazon's, as a channel to plan and measure. Do not treat them as a test to watch. That means ChatGPT ads are not only something to observe. The column says they are working. It says Amazon's ads inside ChatGPT solve a chatbot problem for Amazon. It also covers Walmart accepting Apple Pay. The advice is to plan and measure ad placements inside ChatGPT. That includes Amazon's placements. The column's guidance is to treat them as a channel, not a test. So marketers should give the placements goals and tracking. They should not watch from the sidelines. That is the Stratechery column's position.

That guidance raises a budget question. Before money moves from paid search into ChatGPT Ads, what should marketers know? A Search Engine Journal article lists five questions to answer. The article argues that ChatGPT Ads are not paid search. The format needs its own role, its own budget source, and its own success criteria. Marketers testing ChatGPT Ads should define what the channel is for. They should also define where the money comes from. They should do that before cutting spend on campaigns that already perform. The article's point is that paid search and ChatGPT Ads do different jobs. If the format is not paid search, it should not inherit paid search's rules. It needs a separate purpose, separate funding, and separate measures of success. The five questions are the checklist before any budget shift. They force a clear answer about why the channel exists and what it is replacing. The article says to answer the five questions before cutting spend on campaigns that already perform.

Even with a clear role, ChatGPT Ads may not be a long-term platform bet. A column argues that ChatGPT Ads are a tactical bet, not a platform strategy. The column overlays Statcounter browser share data with SensorTower app data. It says ChatGPT's monthly user growth has flattened since September 2025. It says Gemini keeps climbing. The column concludes ChatGPT Ads can still be worth buying. But it concludes ChatGPT is not the platform to build a long-term plan on. The advice for marketers is to budget ChatGPT Ads as a defined, time-boxed tactic. Keep the platforms they already depend on. That is the column's recommendation. The data it uses comes from Statcounter browser share and SensorTower app data. The growth picture it describes is flat monthly user growth for ChatGPT since September 2025, with Gemini still climbing. The buying advice is separate from the platform advice. Buy the tactic. Do not build the long-term plan on ChatGPT. Budget it as defined and time-boxed. Keep the platforms already in use.

Google is also testing paid formats inside AI answers. Google AI Mode is testing text link ads. The ads look like the normal AI Mode response. They use anchor-text links labeled Sponsored. Those links appear above the AI-generated answer. Marketers should expect sponsored placements to appear in the same surface as AI answers. That gives brands a paid route alongside AI citations. So the same screen that answers a question can also carry a sponsored link. The ad format is built to match the response around it. The label Sponsored is there. The links are anchor text. The placement is above the AI-generated answer. For brands, this means AI Mode can be both an organic citation surface and a paid placement surface. A brand may try to earn a citation in the answer. It may also buy a sponsored link in the same view. Marketers should expect these sponsored placements to sit in the same surface as AI answers. That gives them a paid route next to AI citations. Visibility in AI Mode has more than one lane.

Visibility is also the focus of new manufacturing data. Semrush data on AI search in manufacturing shows which brands get mentioned in AI answers. It shows which brands get cited. It also shows that citations do not necessarily drive traffic. For manufacturing marketers, the advice is to track mentions and citations separately from site traffic. That matters when they judge AI visibility. A brand can be mentioned in an AI answer. A brand can be cited in an AI answer. Citations do not necessarily drive traffic to the site. So manufacturing marketers should not use site traffic as the only measure of AI visibility. They should look at mentions and citations as their own signals. The data separates those outcomes. It shows which brands appear in AI answers and which brands are cited. It also shows that citation and traffic do not always move together. The practical step is to track mentions and citations on their own. Then judge AI visibility with more than one number.

Thanks for listening to Agent Beat. For the full written brief, go to agent-beat.com.