With each new announcement, OpenAI makes the case for why facilitating retail transactions is at the heart of the company’s ChatGPT consumer strategy today.
Yesterday, OpenAI’s Head of ChatGPT Nick Turley announced shopping research in ChatGPT which he described as “a new experience that does the hard part of shopping for you.”
He explained on Linkedin:
“Instead of hunting across dozens of sites, you can describe what you need and get a thoughtful, personalized buyer’s guide in minutes that asks smart clarifying questions and researches deeply across the internet.
Rolling out on web and mobile over the next few days for logged-in ChatGPT users across Plus, Pro, Go, and Free tiers, with nearly unlimited usage through the holidays.”
Read Mr. Turley on LinkedIn. (November 24)
OpenAI’s website offered a look behind the curtain on how shopping research works:
“Shopping research is powered by a version of GPT‑5 mini trained with reinforcement learning specifically for shopping tasks. We trained it to read trusted sites, cite reliable sources, and synthesize information across many sources to produce high-quality product research. We also designed it to be an interactive experience that could update and refine its research in real time—incorporating new constraints and adjusting to feedback on user product preferences—resulting in a response that feels both well-researched and personalized.“
Read: Introducing shopping research in ChatGPT (November 24) – OpenAI
Reaction: Brian Stempeck, Co-founder and CEO at answer engine optimization company Evertune AI, offered a detailed critique of OpenAI’s “Shopping Research” including, “As the model is thinking, it highlights trivia facts, like when the first sunscreen was invented. Have to wonder – is this where the ads go??” Read more on LinkedIn.
From tipsheet: A “free” buyers guide needs ads. Imagine the intent signal bubbling within.
The results from “shopping research” look impressive. However, in my limited testing, it took a few minutes to receive the final output. The immediate return of results of a non-”shopping research” query in ChatGPT was more satisfying due to the speed (the OpenAI graph above says I’m losing accuracy). But, maybe that’s just me. OpenAI’s blog post admits to limitations with the research function at this early stage.
Good for OpenAI to throw it out and test with users. They should do the same thing with ads.
Related: Agentic AI in Retail: How Autonomous Shopping Is Redefining the Customer Journey (November 17) – Bain & Co.
LLMs & CHATBOTS
Developments
- Introducing Claude Opus 4.5; Anthropic says it’s “the best model in the world for coding, agents, and computer use” (November 24) – Anthropic
- “Sam Altman and Jony Ive say they have settled on a design for OpenAI’s device, speaking to Laurene Powell Jobs; Ive says it could debut in ‘less than’ two years.” (November 24) – The Verge
- Amazon to invest up to $50 billion to expand AI and supercomputing infrastructure for US government agencies (November 24) – Amazon
TECH
International search meets the answer engine
Search engine optimization (SEO) firm Previsible claims that OpenAI’s ChatGPT may be competing with Google Search in the U.S., but the international markets tell a different story.
In a post on the Previsible blog, Chilean-based SEO consultant Ana Fernández outlined three key issues:
1. “LLMs use contextual clues + optional IP context, not geo-indexing – Even when an LLM recognizes your IP region, its response still depends on the text of the prompt, rather than a geo-ranked index like Google’s.“
2. “LLMs don’t have ‘market boundaries’, only linguistic and contextual ones Google distinguishes.“
3. “When signals conflict, LLMs fall back to the ‘statistically safe’ answer If the cues aren’t clear, the model predicts the scenario that most commonly fits the tokens.“
She concluded that “As LLMs continue redefining how users find and interpret information, international brands need to prepare for a future where visibility depends on both search engines and AI models.”
Read the post here. (November 18)
From tipsheet: Ms. Fernández is speaking her “book” of course. But she makes a compelling case on where OpenAI may be falling short internationally and speaks to all the ground (opportunity!) OpenAI has yet to cover when it gets to ads in ChatGPT.
RESEARCH
eMarketer: ChatGPT’s “remarkable” reach
Vladimir Hanzlik, Chief Content Officer at research firm eMarketer, wanted to make sure his followers understood how unique OpenAI’s ChatGPT really is.
On LinkedIn, he prefaced a chart on ChatGPT’s reach saying:
“Yes, OpenAI will keep getting hard questions about monetization, Google’s latest Gemini model, or Microsoft and NVIDIA’s investing in OpenAI’s competitor Anthropic, but ChatGPT’s ability to reach consumers remains remarkable.”
Read more on LinkedIn. (November 24)
PROTOCOLS
AdCP will level the playing field
After emerging from his marketing “sweat lodge” over the weekend, Rembrand CMO Cory Treffiletti had insights to share about Ad Context Protocol (AdCP) on LinkedIn.
Mr. Treffiletti is an AdCP believer as is Rembrand CEO Omar Tawakol (here).
Channeling a past life in the agency world, Treffiletti explained yesterday:
“What I thought about was just how easy it will be to be a media planner when AdCP hooks are enabled, and you can do basic planning with larger platforms through generative AI solutions.
I also realized how easy it is to create these platforms, and how the large agencies no longer have a competitive advantage when it comes to media planning and buying.”
He offered his three AdCP takeaways:
- “AdCP is very important for creating a level playing field.”
- “App creation is super simple these days (If I can do it, anyone can do it).”
- “Competitive advantage lies in advanced strategy, and not basic planning.”
Read more on LinkedIn. (November 24)
BRANDS
Opinion: Automation and discernment are next
In an op-ed on AdExchanger, Anastasia Leng, founder and CEO of creative analytics firm CreativeX, made the case for the human in the automated advertising future.
As with any attribution analytics firm today, she emphasized the importance of measuring outcomes. But in this case, the outcome was the original creative idea generated by the marketer.
Ms. Leng called it moving from outputs to ideas. She concluded:
“That’s the paradox of the self-driving age: The more we automate, the more human discernment matters. The promise of automation was to make marketing smarter. The paradox is that it’s made judgment more valuable than ever.”
Read more on AdExchanger. (November 24)
LLMs & CHATBOTS
GEO (or AEO) adds inline images
Late last week, OpenAI quietly rolled out more inline images within responses in ChatGPT.
For generative engine optimization (GEO) companies and marketers, that means another lever for placing a brand or product listing in the answer engine.
Read ChatGPT “Help.” (November 21)
RESEARCH
Maximizing subscription + ads yield
“Fascinating paper from researchers at Pandora that explores how revenue can be optimized across subscriptions and advertising by personalizing ad load.”
Read: “Personalizing Ad Load to Optimize Subscription and Ad Revenues: Product Strategies Constructed from Experiments on Pandora” (2024) – SSRN
PEOPLE MOVES
Hiring, layoffs
- Mark Read Lands First Post-WPP Role as Kantar Media Chairman (November 24) – Adweek
- Layoffs at Yahoo DSP (November 24) – Kameron Canbaz, Yahoo DSP on LinkedIn
MORE
- Omnicom and Interpublic Receive Unconditional Clearance from the European Commission (November 24) – OmnicomGroup
- Amazon quietly blocks more of OpenAI’s ChatGPT web crawlers from accessing its site (November 21) – Modern Retail
- Google Performance Max gets direct video uploads (November 24) – Search Engine Land
- “We raised $21M Series A to help brands win in AI search” (November 18) – Peec AI