LLMs & CHATBOTS
Usage vs. training deals
In reviewing publishers’ licensing deals with large language models (LLMs), Digiday’s Jessica Davies focused on the usage deals a.k.a. “AI grounding” or Retrieval-Augmented Generation (RAG) in an article on Friday.
Unlike training deals where the LLMs visit a publisher site once and then references that data for chatbot responses, with RAG, LLMs dynamically fetch real-time information from publishers to generate responses. “Data freshness matters” to LLMs is one takeaway for these deals. Another is that the usage deals are “fairer” to publishers and presumably they make more money whenever their content is cited and linked to.
Read more in Digiday. (August 22)
From tipsheet: LLMs want more access and, ideally, real-time access to publisher sites because it makes their answer engine’s content better.
But do publishers really want to be paid on a lump sum or per impression/usage basis? It still feels that it’s hard to say which is the better model at this point. Publishers appear to be attracted by the potential of being paid every time an LLM crawler accesses their sites and then uses them in an answer. That sounds fine. But, it’s all in the math. Some form of AI-real-time-bidding (or AIRTB) feels inevitable.
On the LLM crawlers side, real-time crawling makes sense in terms of understanding the latest and greatest on behalf the answer engine user. But, the compute required to keep track of the marketplace of crawling seems huge and potentially very expensive, if not prohibitive, at scale. Of course, that’s what they said about RTB in its infancy, too.
Related:
- “IAB Tech Lab is launching our new Content Monetization Protocols (CoMP) for AI Working Group to help Publishers monetize content in a world of LLMs and AI models. Are you interested in joining?” (August 20) – Anthony Katsur, CEO, IAB Tech Lab on LinkedIn
- Announcing “Content Monetization Protocols (CoMP) for AI” Working Group – IAB Tech Lab
- Publishers Explore New Ways to Get Paid When AI Uses Their Work (August 22) – Adweek
LLMs & CHATBOTS
Parts of an OpenAI deal
Jonathan Roberts, Chief Innovation Officer, People, Inc. (formerly Dotdash Meredith), spoke to Marketecture about publishers licensing to LLMs and AI transformation. He described his company’s multi-year deal with OpenAI in May of 2024.
“It’s got three parts to it.
One is payment for training…
And then the second is that every time our content is used, it requires meaningful attribution. So we actually got citation into the partnership…
And then the third piece was co-development.
They helped us build our new tech – so D/Cipher (People’s ad targeting tool)…”
Hear more on the Apple Podcasts app. (August 22)
From tipsheet: This is a good podcast where Mr. Roberts appears to accept that AI transformation is inevitable for a big publisher such as People. In fact, he says he feels “a lot more optimistic” about his company’s prospects in an AI environment than he did as recently as May. But, that doesn’t mean his company isn’t aggressive about protecting its intellectual property.
LLMs & CHATBOTS
That MIT report
Mainstream media gobbled up snippets of a MIT-funded research paper last week on AI investment.
Advertising and AI – and the billions being made by companies such as Google, Meta, Amazon, Microsoft and ad technology-related companies – did not fit into the “pilots” rubric used by researchers.
- Original MSM article: MIT report: 95% of generative AI pilots at companies are failing (August 18) – Fortune
- More MSM coverage: MIT study on AI profits rattles tech investors (August 21) – Axios
- Reviewing the media coverage: MIT report misunderstood: Shadow AI economy booms while headlines cry failure (August 21) – VentureBeat
- The paper: The GenAI Divide: State of an AI in Business 2025 (PDF) – MIT paper on 3rd party website, MLQ
AGENCIES
AI-enabled conversions
“Why agencies are optimistic about conversions despite AI search traffic woes” (August 22) – Ad Age (subscription)
SELL-SIDE
Trumpeting AI-enabled Answers
On Friday, Reddit CRO Mike Romoff (formerly of LinkedIn) was promoting on LinkedIn a new Fast Company article about his company’s AI-enabled “Answers” product:
“If you haven’t tried Reddit.com/answers yet you’re missing out. Go ahead and check it out. Yes now. It’s ok, I’ll wait….Fast Company says it well: “Reddit provides happy evidence that doubling down on humans can be a viable business.”
The company’s own AI answer engine product is driving Reddit user engagement and, consequently, its ad business. Of course the Reddit CRO is happy!
How is it different from a Google search engine result? Web Pro News explained, “[Reddit Answers] aggregates and summarizes user-generated posts and comments, providing a distilled view of real-world experiences rather than SEO-optimized content that can feel detached and promotional.”
Read: “Reddit—and a dash of AI—do what Google and ChatGPT can’t” (August 22) – Fast Company (subscription)
More: “Reddit Answers – Currently in Beta” (August 1) – Reddit
From tipsheet: How many times have you gone to Google and tried to force a result that comes from the Reddit community? Reddit’s AI-enabled answer engine replicates that need. Mobile Dev Memo’s Eric Seufert dicussed the Reddit “double dip” last week.
SELL-SIDE
Good crawler attributes
Martin Splitt from Google Developer Relations participated in an impromptu web crawling discussion on Bluesky last week and provided this list of best-practice attributes:
“support http/2
declare identity in the user agent
respect robots.txt
backoff if the server slows
follow caching directives*
reasonable retry mechanisms
follow redirects
handle errors gracefully*
Google’s Martin Splitt and Gary Illyes talked “Good Web Crawler Attributes” according to Search Engine Roundtable. Read more. (August 22)
This is very relevant to Google’s own crawling conundrum: crawling for its search engine versus crawling on behalf of AI Overviews. As the company tries to shift to an AI-enabled future, it’s also trying to preserve “legacy” (my word) ad revenue on the search engine which also preserves the value exchange with publishers.
Reportedly, the company’s crawlers can only be blocked on an “all or nothing” basis. In other words, a publisher can’t block the AI Overviews crawler without blocking the search engine crawler.
TECH
TikTok ads and AI
TikTok’s AI ad system known as “GMV Max” – which is intended to “to maximize Gross Merchandise Value (GMV) by using artificial intelligence to optimize campaigns” -appears to be creating consternation with marketers who see another “black box” getting in their way.
Meanwhile, the company appears to be laying off content moderators in the UK as AI takes over some of the duties.
- TikTok is mandating that Shop brands use a new AI ad tool starting September 1, and some aren’t happy. (August 21) – Business Insider
- TikTok puts hundreds of UK content moderator jobs at risk, AI moderation may be next (August 22) – BBC
Related:
TECH
Explaining Google Search and AI
Google executives Nick Fox, SVP of Knowledge and Information, and Liz Reid, VP of Search at Google joined “The Big Technology” podcast host Alex Kantrowitz for an hour-long conversation about AI and, mostly, the evolution of search.
The podcast, which was published last Thursday, was recorded around the time Ms. Reid was defending Google search and its referrals to publishers in a Google blog post.
Several podcast highlights:
- GEO is new: Mr. Fox admitted that he had not heard of the GEO acronym, or Generative Engine Optimization, until the day of the podcast: “I think I heard GEO for the first time today. Actually, I hadn’t heard that term before.” This speaks to how fast and new the AI search space is moving.
- SEO, AEO, GEO – Optimization is the same: In speaking to the proper way to optimize for placing content in an answer engine (AEO, GEO and even SEO), Ms. Reid said what Google has always said about search optimization, “And so I think if marketers have to just focus on building content for people, that’s awesome.” Mr. Fox added with more of a publisher focus, “Our general advice to websites is the way you optimize for generative AI is actually the same” as Google Search.
- Open web and AI: Mr. Fox emphasized Google “deeply care[s]” about the open web and is aware of publishers dependency on traffic from Search. But, he also said Google has to lean in to AI: “We have to evolve to where users are, where users are going, what the technology can do(…) Our approach is to do it in a way that enables the web to thrive through that moment.”
- Four search momentum drivers: Mr. Fox also discussed the phenomenon of how Google Search queries and revenues are growing which he attributed to generative AI for four reasons: 1) Users realize they can ask more questions and get answers; 2) More queries driven by the AI Overviews format means more views for advertisers in the search engine result pages below; 3) Generative AI is being used in Google’s search ad products which helps drive ROI for advertisers; 4) Increased specificity of the search queries due to gen AI “leads to more valuable traffic to advertisers.”
Listen to the podcast on the Apple Podcasts app. (August 21)
TECH
AI in the video
Kuaishou, China’s No. 2 short-video platform, behind Douyin, TiKTok’s sister app in China, reported Q2 2025 profit that beat analyst expectations for its ad business and AI monetization.
Topline revenue reached nearly $5 billion USD.
Read more in The Wall Street Journal. (August 21)
See Kuaishou’s English language website.
PROMPT
How will vertical search be affected by AI?
Response from Perplexity AI:
AI is dramatically reshaping vertical search, driving a fundamental shift in how users discover and interact with specialized information, products, and expertise within industry-specific platforms…
MORE
- Opinion: How AI Search Is Reshaping B2B Marketing (August 22) – Forbes
- Critics Say The Trade Desk Is Forcing Kokai Adoption, But Apparently It’s Up To Agencies (August 22) – AdExchanger
- Google Ads rolls out Asset Studio beta (August 22) – Search Engine Land
- Opinion: Musk’s Near $100B OpenAI Bid Could Complicate Advertising – MediaPost
- Nvidia earnings set to test AI trade with stocks near record highs: What to watch this week (August 24) – Yahoo! Finance