OpenAI Launches ChatGPT Agent

OpenAI Chief Product Officer Kevin Weil said yesterday on LinkedIn:

• “Announcing ChatGPT agent: a powerful new agent that can use a computer, browse the web, write code, use a terminal, write reports, create images, edit spreadsheets, and even create slides for you.”

Read more from Mr. Weil on LinkedIn.

TechCrunch reported “that users can tap ChatGPT agent to ‘plan and buy ingredients to make Japanese breakfast for four,’ as well as ‘analyze three competitors and create a slide deck,’” according to OpenAI.

Read about it:

• “OpenAI launches a general purpose agent in ChatGPT” (July 17) – TechCrunch

Is there an ads and marketing application here? Gotta be. How about planning and buying some ads instead of a Japanese breakfast?

OpenAI’s X account discussed the broader potential: “ChatGPT agent’s capabilities are reflected in its state-of-the-art performance on academic and real-world task evaluations, like data modeling, spreadsheet editing, and investment banking.” See it now.

BRANDS

The move to “outcome”pricing

Adweek’s Trisha Ostwal reported Wednesday that pricing for ad serving may be changing – from CPMs to a (agentic?) flat fee.

She teased her article on LinkedIn:

“Another ad tech domino falls. Clinch, an AI agent-powered ad platform, is ditching CPM-based pricing—long considered the industry’s default—for a flat monthly fee that covers unlimited ad serving across all channels. Why it matters: Brands like Hyundai, Albertsons, Coca-Cola, and Kroger pushed back on pay-per-impression models and asked for more predictable pricing. Since introducing the model earlier this year, Clinch has landed 10 new Fortune 500 clients and estimates 30–50% savings per client versus CPMs.”

Flat fee… another baby step closer to real outcomes in the age of AI?

Read: “Agentic Adtech Firm Clinch Scraps CPMs for Monthly Subscription Pricing” (July 16) – Adweek

BRANDS

Curation and outcomes, and now AI

In a questionable open web, there is little evidence yet that ‘curation’ and accountability for ‘outcomes’ are effective and improve transparency. So what needs to change?

Read: Op-ed – “Curation and outcomes: Step in the right direction or another false dawn?” – Nick Manning, co-founder of MG OMD, The Media Leader UK (July 17)

AGENCIES

About agency earnings and AI

• Publicis‘ CEO dismisses Meta threat, raises yearly growth guidance (July 17) – Reuters
• Omnicom Posts Lower Earnings Amid Interpublic Acquisition (July 15) – The Wall Street Journal

Related: Inside The ANA’s New Legal Gen AI Template For Advertisers And Agencies (July 17) – AdExchanger

SELL-SIDE

Investment, division in creator economy

Newsletter platform Substack announced yesterday that it has raised $100 million in Series C financing to service its newsletter creator clientele.

On the Substack blog, a trio of Substack executives announced the raise and “slapped” the collective faces of Instagram, TikTok, X and others:

“…this time of flux also holds tremendous opportunity. A growing number of people are navigating the chaos by choosing independence. Audiences are investing their attention and money in what they value, not just what addicts them. Creators are building livelihoods based on trust, quality, and creative freedom. They know the future belongs to those who build it.”

Read more.

Thought bubble: Last I checked, social media apps are a critical tool for distribution with newsletter writers. So, bring on the addiction! Nevertheless, the content creation community gets a boost with the latest Substack investment which includes support from techno-optimist venture firm Andreessen Horowitz.

SELL-SIDE

Analysis: Monetizing AI for content creators

On Wednesday, Stratechery’s Ben Thompson reviewed Cloudflare’s recent pay-per-crawl announcement and saw the future of content – it’s a marketplace. But some may have to swallow a bitter pill – let Google Search thrive.

ICYMI, Cloudflare, one of the web’s largest content delivery networks, now blocks all AI bot crawlers by “default” on behalf of publishers rather than an “easy button” that Cloudflare’s publisher clients had to initiate and which had already existed for over a year.

Mr. Thompson noted Wednesday (subscription) the important difference between some of Google’s AI bot crawling and all other bots which is a function of details unearthed during recent antitrust proceedings. Mr. Thompson explained:

“Google’s Advantage”

“There is one important exception to these new defaults: Google, which has two crawlers. Googlebot crawls the web for Google search, while Google-Extended crawls the web to capture data for Gemini. What is critical to understand, however, is that data for Google Search AI products — including AI Overviews and AI Mode (i.e. the search funnel) — is gathered by Googlebot; that means that if you want your website to show up in Google Search you have no choice but to have that data also be used by any AI products that are under the Search umbrella.”

Mr. Thompson said he sees potential for a new content marketplace in a publisher world that is “rapidly being overwhelmed by AI” and described how it would work in May. Here’s a snippet:

“First, the protocol layer should have a mechanism for payments via digital currency, i.e. stablecoins. Second, AI providers like ChatGPT should build an auction mechanism that pays out content sources based on the frequency with which they are cited in AI answers. The result would be a new universe of creators who will be incentivized to produce high quality content that is more likely to be useful to AI, competing in a marketplace a la the open web; indeed, this would be the new open web, but one that operates at even greater scale than the current web given the fact that human attention is a scarce resource, while the number of potential agents is infinite.”

Thompson’s “move to a content marketplace” point-of-view syncs with Cloudflare CEO Matthew Prince’s suggestion about a content marketplace just after he announced the “default on” setting and pay-per-crawl future for AI bots.

Prince wrote on July 1:

“Next, we’ll work on a marketplace where content creators and AI companies, large and small, can come together. Traffic was always a poor proxy for value. We think we can do better. Let me explain.“

“Imagine an AI engine like a block of swiss cheese. New, original content that fills one of the holes in the AI engine’s block of cheese is more valuable than repetitive, low-value content that unfortunately dominates much of the web today…“

Bottom line:

• Messrs. Thompson and Prince are making the case that the content marketplace is the solution for content in the age of AI – and AI bot crawling. (i.e. the new payment system)

• Thompson and Prince are also making the case that you can’t threaten publishers’ Google Search referral traffic.

• Finally, Thompson, Prince and Google are making the case that using AI products (AI Overviews, AI Mode, etc. – known as “search generative experiences”) in the search results is important for maintaining critical search referral traffic to publishers. (i.e. the old payment system)

Policy pro alert: From here… Microsoft, Oracle, Meta, OpenAI, Anthropic and other LLM makers, in theory, would take issue with the stance by Cloudflare and Google, in particular.

Will policy teams get busy in DC? We’ll see.

Nevertheless, we’re starting to transition into the AI age.

SELL-SIDE

Pay-per-whatever

This marketplace solution described by Stratechery and Cloudflare is just the high-level.

As Cloudflare engineers theorized on July 1, are payments negotiated on a pay-per-crawl basis, pay-per query basis, pay-per inference basis?…

How do you introduce granular licenses at internet scale, whether for training, inference, search, or something entirely new?

The true potential of pay-per-crawl may emerge in an agentic world.

So much to come. But it’s yet-another application of the agentic world that will be tied to the full funnel of advertising and marketing in the future.

One more thought on Cloudflare: Cloudflare does not control all the web’s traffic. According to its own 2023 data – Cloudflare controls 16% of all internet traffic. So:

• This marketplace is going to be great lead gen for Cloudflare and its publisher tools.
• Opportunities remain for all the other CDNs such as Fastly and even Google.
• And startups targeting pay-per-whatever – such as TollBit – have opportunity waiting for them too.

More: “IAB Tech Lab pitches plan to help publishers gain control of LLM scraping” (July 16) – Digiday

TECH

Seufert: Google Gambit

On Wednesday, Mobile Dev Memo’s Eric Seufert analyzed recent data from investment banking firm Oppenheimer & Co. which showed that “of users familiar with AI Mode and who pay for ChatGPT, 82% find AI Mode more helpful than Google Search and 75% find AI Mode more helpful than ChatGPT.”

Seufert ties the data to what he calls the “Google Gambit” (Seufert’s post 1, post 2, and Wednesday’s post).

An essential part of this “gambit” by Google is what Seufert says is moving is search from a “distribution mechanism” to “engagement sync.” The implications are that Google wants users spending more time on their product – bad for publishers potentially.

Mr. Seufert wrote on Wednesday about the challenge for publishers:

• “This puts scaled web publishers in a difficult position: if they wish to retain the inbound Google traffic on which their businesses have been built and fundamentally depend, they may have to meet Google on its new terms: to make their data available for training in exchange for visibility in AI Overviews.”

Read more from Eric Seufert on Mobile Dev Memo (July 16).

Thought bubble: Mr. Seufert is more “glass-half empty” on the Google Search and AI strategy than Stratechery’s Ben Thompson who appears to see a new day dawning for a content marketplace and pay-per-crawl in the age of AI. Both are digging into the depths of the future of advertising, marketing and media like few others these days.

Time spent vs. retention

OpenAI has specifically said they are not optimizing for time spent, i.e. engagement.

On a July 1 OpenAI podcast, Head of ChatGPT Nick Turley said, “So for us, ‘time spent,’ – it’s very much not the thing we optimize for. We do care about your long-term retention, because we do think that’s a sign of value.”

If Google is optimizing for time spent (or engagement) and OpenAI is not, will that point of differentiation lead to the ultimate winner/loser in the chatbot experience?

TECH

AWS event: generative AI creative

Along with Thompson and Seufert, New Street Research’s Dan Salmon continues to pen astute coverage of publicly-traded companies – such as Google and Amazon – from a Wall Street perspective.

On Wednesday, Mr. Salmon attended the big Amazon Web Services (AWS) event in NYC and reported back to clients:

• Agentic AI was the biggest theme of the day with all vendors supporting
• The 2nd biggest theme was a new focus on the AWS Nova family of models
• Bedrock adds AgentCore in line with the #1 theme: also supporting cloud migrations
• Anthropic was the belle of the ball
• Heavy presence from NVDA, less so for Trainium and Inferentia
• AWS Clean Rooms and gen AI ad creative in focus for digital advertising

We came away with more conviction in the breadth of AWS AI services.

Salmon sees AWS revenue continuing to accelerate well-above Wall Street’s current estimates in the second-half of 2025 and into 2026.

TECH

AI Startups: Consulting to Platform

“Today, we become MadConnect. What started as a consultancy solving complex marketing challenges has evolved into something much bigger: a platform. More specifically, the infrastructure layer that powers modern, AI-ready marketing.”

CEO Bob Walczak of MadConnect on LinkedIn Wednesday.

PROMPT

Prompt: Discuss the strategy for effectively adding advertising to an LLM chatbot

Response from OpenAI’s ChatGPT:

Adding advertising to an LLM chatbot—effectively and sustainably—requires balancing monetization with user trust, utility, and brand safety. Below is a strategic breakdown for doing this well:

#1 – Define the Monetization Model

There are three primary advertising models for LLM chatbots:

• Sponsored Answers – Brands pay to have their products or services included in responses. This is akin to “sponsored search results” but within conversation.
• Conversational Product Placement – Ads are naturally embedded into replies when contextually relevant (e.g., travel chatbot suggesting a hotel chain).
• Zero-Click Commerce / Affiliate Links – The LLM suggests a product or service with a direct link (affiliate or otherwise), driving instant purchase without needing a traditional click-through funnel.

#2 – Ad Targeting and Contextual Relevance…

Read more on tipsheet.

---

MORE

• “OK, so it’s agentic. But what else does it do for me?…” (July 17) – Scope3 CEO Brian O’Kelley on LinkedIn
• After 20 years chasing data efficiency, ad tech is finally prioritizing creative again (July 14) – State of Streaming
• Google Search adds Gemini 2.5 Pro, AI Deep Search, and agentic local calling (July 16) – Search Engine Land
• Oracle Is No Longer AI’s Dark Horse (July 16) – The Wall Street Journal