Taboola is rolling out the beta of Realize+ positioned as an “agentic” system that continuously plans, executes, and optimizes campaigns for performance advertisers.
The ad tech company originally launched Realize for performance advertisers in February 2025. The platform, with particular focus on open web inventory, now centers on a closed-loop model — dynamically allocating budget and iterating creative and targeting in real time — and is aimed at replicating the automation seen in Google’s Performance Max and Meta’s Advantage+.
Separately, Taboola is integrating Claude as a “skill,” giving agencies and advertisers a conversational interface to set up and manage campaigns within the system.
CEO Adam Singolda said on LinkedIn yesterday:
“Today, we’re launching a Claude Skill for agencies and advertisers. A shift from ‘Ad Ops’ to ‘Agentic Ops.’
By plugging Claude directly into our agentic AI buying platform, we’re enabling a future where goals execute themselves across the open web.
From managing campaigns to managing outcomes.
AI agents don’t just assist anymore, they operate.”
Read it on LinkedIn. (April 23)
Promising more Claude skills to-come, the press release explained: “Realize+ has spent several months in beta. Stemming from positive early results, phase 2 of the beta will now be expanding in Q2 with improved models, while also featuring an extensive roadmap throughout 2026.”
Read: Taboola Launches Realize+, an Agentic AI System Turning Advertiser Goals Into Outcomes and Expands Agentic Capabilities by Opening its Platform to Claude Skills (April 23) – Taboola
AI Video: “Realize+ is Here: Outfox Your Competition with Agentic Advertising” (April 23) – Taboola on YouTube
From tipsheet: The question is who owns the agent over time: the model layer or the media layer. In other words, does Claude decide where and how dollars flow across platforms or does Taboola keep that control inside its own optimization engine?
And so, I reached out to Taboola and asked that question.
Here is the Taboola spokesperson’s response:
“There are two parts to what we announced today.
First, within our advertising platform Realize, we launched Realize+, an agentic experience that automates campaign setup, optimization, and scaling end-to-end. Advertisers can choose Realize for more hands-on control, or Realize+ for full automation, similar in spirit to Google Performance Max or Meta Advantage+.
Second, we made Realize and Realize+ available as a Claude Skill, because advertisers and agencies are increasingly using AI agents like Claude to coordinate marketing work across multiple channels.
On your question of control: Claude is one interface and orchestration layer an advertiser can use to set goals and workflows. Execution and optimization is then on Taboola – how budgets are deployed, bid, paced, and scaled, those things stay inside Realize, powered by our bidding models, first-party performance data, and distribution across publishers, OEMs, and apps.
The model layer is getting very good at intent, orchestration, and cross-platform coordination. But performance inside any given media environment depends on signals and supply the agent doesn’t see – which publishers are converting for which audiences right now, how inventory is pricing, what creative is working where. That’s the layer tech platforms own, and where the outcomes actually get produced.
We think agents help advertisers operate across platforms, and each media platform continues to own performance, optimization, and outcomes within its environment. Claude Skill is how we plug into that world to start.”
LLMs & CHATBOTS
Developments
- Introducing GPT‑5.5 (April 23) – OpenAI
- Copilot’s agentic capabilities in Word, Excel, and PowerPoint are generally available (April 22) – Microsoft 365 blog
- Elon Musk lays out Terafab AI chip project plan (April 23) – Reuters (subscription)
MARKETING
Unpacking the ‘Frontier CMO ethos’
Google is leaning into the idea that the latest AI ‘frontier models’ are reshaping the role of the CMO.
As part of a Google promotional podcast series “Frontier CMO”, L’Oréal Chief Digital & Marketing Officer Asmita Dubey speaks with Josh Spanier, Google’s VP AI & Marketing Strategy.
When a CMO like Dubey speaks, the industry listens.
Spanier called Dubey the ‘Frontier CMO blueprint’ in a LinkedIn preview of the interview:
“L’Oréal may be a century-old brand, but it’s committed to moving at the speed of a startup. I sat down with Asmita Dubey to discuss how she helps drive over $40 billion in revenue by bridging the gap between hard science and the ephemeral magic of beauty.
We dive into upskilling 70,000 employees on AI and why she’s vertically re-bundling marketing silos.”
Read more on LinkedIn. (April 23)
In the article, Dubey breaks down AI transformation for her global brand into three pillars:
- Re-bundling
- Scaling humans
- Innovation markets
On “scaling humans,” Google summarized Ms. Dubey on LinkedIn:
“CMO move: Stop looking for the perfect AI tool and start building a Skills University. Your competitive advantage in 2026 isn’t owning the best LLM; it’s having a workforce that knows how to flex the dual muscle of math and magic.”
Read: “Math, magic, and Asmita Dubey’s $40 billion glow up” (April 23) – Think With Google on LinkedIn
Listen: “Frontier CMO” podcast on YouTube
From tipsheet: The overall point here is that owning, or using, the best model won’t matter. The winners will be the orgs that can actually optimize for AI by using it at scale, across functions, with humans trained to push beyond the just-OK outputs.
CONNECTED TV
Hot TV takes
- Here Come the TV Outcome Wars (April 23) – Mike Shields on his Next in Media Substack
- Consumer trends in 2026: how consumers are redefining TV – Jason Fairchild, CEO, Pinterest’s tvScientific on his Beehiiv blog
LLMs & CHATBOTS
Inside the OpenAI vs. Anthropic brand war
Ad Age chief technology reporter Garett Sloane frames the emerging brand battle between OpenAI and Anthropic as both companies ramp up marketing alongside model development.
Sloane previewed the piece on LinkedIn:
“Here is the brand story within the AI story. With all the focus on new models and ad businesses — ‘Mythos is too powerful’ — and AI in marketing, we forget the AI brands are marketing, a lot. Here is a reprise of the brand battle at Ad Age. This one between OpenAI and Anthropic, a communication war that gets more heated by the day, and now there is spending to it.
-
- Sensor Tower tells us that Gen AI as a category makes up 25% of all software marketing spend, up from 4% in 2024. So that just outlines the contours.
- OpenAI and Anthropic really kicked this off around Super Bowl, and they both are among the top 85 spenders in U.S. advertising right now.
- So what’s the strategy? I spoke with Mother, Dentsu Creative, and others to get the context of how AI is breaking into advertising mindset.”
Read it on LinkedIn. (April 23)
More: OpenAI vs. Anthropic — how the AI players seek to dominate the market (April 23) – Ad Age (subscription)
From tipsheet: The AI wars are heading to Cannes. First comes brand spend, then come the sponsored mega-yachts.
AGENTS
Q&A: TripleLift launches TL Spark agent
On Tuesday, creative sell-side platform TripleLift launched TL Spark — an agent that acts as an intelligence layer for campaign execution.
TL Spark’s core components reach across curation, creative, audiences, measurement and optimization, according to the company.
A press release explained, “By coordinating these elements rather than leaving them in silos, TL Spark learns from multivariable performance patterns — how creative, inventory, audiences, and measurement interact — and feeds those insights back into planning and optimization.”
Read: “TripleLift Launches TL Spark, Introducing an Intelligence Layer for Outcome-Driven Advertising” (April 21) – TripleLift
TripleLift CEO David Helmreich answered follow-up questions from tipsheet via email.
tipsheet: “What is the next step for TL Spark as you move toward more autonomous agentic execution?”
David Helmreich, CEO, TripleLift: “The next step is earning the right to act. Autonomous systems need guardrails before they get autonomy, and that is where we are investing: feedback loops, performance transparency, and governance infrastructure that make it responsible to let the system do more.”
“The vision is a coordinated set of specialized agents handling supply, audiences, creative, measurement, and optimization together, continuously, with an orchestration layer tying them toward a campaign objective without requiring human intervention at every step. We are closer to that than most people expect, particularly as industry frameworks for agentic advertising mature on both the sell-side infrastructure and buy-side communication layers.”
“TL Spark is being built to be part of an open, interoperable agentic ecosystem. The standards are still evolving but the direction is clear, and we think being an early participant in shaping them is exactly where TripleLift should be.”
Additional questions include:
- What decisions can TL Spark make autonomously today and what still requires human input?
- Does TL Spark directly execute campaign changes, or does it surface recommendations for traders to act on?
- When you say ‘creative transformation,’ what is actually generated by the system versus provided by the advertiser?
- Are measurement signals directly driving in-flight decisions in real time, or are they used for post-analysis and planning?
Read: “Q&A: TripleLift launches TL Spark agent” (April 8) – tipsheet
TECH
Seufert: Meta optimizing advertiser properties
Previewing his latest piece on Mobile Dev Memo, analyst Eric Seufert posted on LinkedIn about Meta pixel and CAPI announcements:
“Last week, Meta introduced automation updates to its two keystone data ingestion features for advertisers: its pixel and its Conversions API (CAPI). These updates are less significant as product enhancements than as evidence of Meta’s ambition to expand its optimization surface to advertisers’ properties themselves.”
Read LinkedIn. (April 23)
More:
- “Meta’s AI pixel and the vanishing ad-product divide” (April 23) – Eric Seufert on Mobile Dev Memo (subscription)
- Removing Technical Barriers to Help Businesses of All Sizes Get More From Their Ads (April 15) – Meta
- Set up the Meta Pixel and Meta Conversions API for ad campaigns – Meta Blueprint
From tipsheet: Pixels used to measure performance. Now they feed the AI machine.
AGENTS
Agentic Ad Management Protocols update
“AAMP 2.0 is here!
With Pricing and Negotiation Engines that preserve existing rate agreements, a Deals Library, Media Kit Management, and more, AAMP 2.0 introduces Buyer and Seller Agent SDKs that are ready for business.
Oh, and we’ve also operationalized Agentic Audiences—so you can start adopting this powerful, privacy-preserving specification today.” – IAB Tech Lab on LinkedIn (April 23)
Read:
- AAMP 2.0 Release Brings Transaction-ready Buyer and Seller Agent SDKs (April 23) – IAB Tech Lab
- AAMP (Agentic Advertising Management Protocols) (updated April 23) – IAB Tech Lab
TECH
Ad tech veterans launch intel platform
Ad tech veterans Jay Friedman and Danilo Tauro announced their new company yesterday, CartographAI, focused on creating a vendor intelligence platform for advertisers and agencies.
Promising neutrality with the platform’s intel, Mr. Friedman explained on LinkedIn:
- “There’s no pay-to-play and vendors can’t pay to rank higher. Any vendor can be listed and assessed at no charge.”
- “Brand/Agency access is free forever. You can’t rely on expensive reports that offer partial coverage and are only updated once/year in an industry that changes weekly.”
- “We are human expert-led. AI-assisted? Of course. This means we deliver structured, expert-led assessments across 40+ tool categories, scored on what buyers actually hire these tools to do.”
Read more on LinkedIn. (April 23)
Regarding the platform’s commercial model, Friedman told tipsheet in email yesterday: “We’re not as concerned with making money these first few months as we are shifting the paradigm and delivering over-the-top value. Once we do that, I’m sure we’ll figure out the best way to charge (but it won’t be to buyers – that’s free forever).”
More: “This Startup Wants to Fix the Slog of Ad tech Vendor Selection With a AI-Supported Matchmaking” (April 23) – Adweek
From tipsheet: In the age of AI, the bottleneck isn’t execution — it’s selection. Models can run campaigns, generate creative, even optimize spend. But they still need to decide where to spend and which vendors to trust. CartographAI wants to sit at that crossroads.
MORE
- ChatGPT Ads Have Begun Showing Up For Logged-Out Users (April 23) – AdExchanger
- Google’s Liz Reid on AI search changes, query shifts, and AI slop (April 23) – Search Engine Land
- Why AdCP was built in public: to make a working standard, not standardize what works (April 22) – Scope3 CEO Brian O’Kelley on his BOK on Ads (and Climate) blog
- Why Gigi Is Going Headless – Adam Epstein, CEO, Gigi on his Substack
- MiQ Report Studying 53 Million Households Shows the Traditional Path to Purchase is Dead (April 22) – press release