Unity sees AI results, shifts strategy

John Ebbert

Unity and Vector

Unity pre-released Q1 2026 earnings yesterday and said that its Vector AI engine — Unity’s AI-enabled prediction and optimization engine — helped drive strategic 48% year-over-year growth.

“Strategic” revenue excludes Unity’s ad network and publishing businesses.

The company also announced it is pivoting away from the lower-margin ad network and publishing businesses toward its optimization layer (Vector AI) that optimizes ads and monetization.

Unity’s ironSource Ads Network and mobile game publishing platform Supersonic will be exited.

CNBC’s Jon Fortt wrote on LinkedIn:

“Unity CEO Matthew Bromberg today announced preliminary Q1 2026 results beating guidance across the board: revenue $505–$508M vs. $480–$490M guided, EBITDA up 58% YoY at ~26% margin. Unity is exiting ironSource Ads Network and Supersonic, revealing strategic ad revenue growing 48% YoY, powered by Vector AI’s fourth straight quarter of mid-teens sequential growth. Bromberg says margin expansion is the new normal, not a one-time boost.”

Read LinkedIn. (March 26)

More: “Unity Releases Preliminary First Quarter Results Exceeding Guidance; Will Enhance Growth and Profitability by Exiting Non-Strategic Ad Businesses” (March 26) – Unity

Video: Interview with Matthew Bromberg, CEO, Unity (March 26) – CNBC

From tipsheet: Unity has two core parts to its business:

  • “Create” – Subscriptions to its 3D development tools and engine.
  • “Grow” – Ads and monetization of third-party mobile apps powered by Vector.

Unity’s exit from mobile publishing (Supersonic) and its legacy ad network (ironSource) echoes AppLovin’s earlier, gradual pivot away from owned games toward its Axon AI ad platform and ‘pure advertising.’ (AdExchanger, February 2025)

In both cases, owning games is giving way to owning the models that monetize them.

LLMs & CHATBOTS

Developments

LLMs & CHATBOTS

OpenAI talks ChatGPT ad revenue

After six weeks of its US-only ads pilot, OpenAI gave an update to The Information’s Stephanie Palazzo on how the test phase is progressing.

Palazzo reported yesterday:

“OpenAI has surpassed $100 million in annualized revenue from its ChatGPT ads business, six weeks after the pilot was announced, according to a spokesperson. That revenue has been generated from the less than 20% of U.S.-based ChatGPT Free and Go users who are shown ads on a daily basis today, though around 85% of Free and Go users are eligible to see ads, the spokesperson said.

OpenAI has expanded to more than 600 advertisers, and is on track to launch self-serve advertiser access in April, they said. The company is currently focused on improving ad relevance, with less than 7% of ads rated by users as ‘low relevance,’ as well as ensuring that ads don’t negatively impact user trust, the spokesperson said.”

Read: “OpenAI Surpasses $100 Million Annualized Revenue From Ads Pilot” (March 26) – The Information (subscription)

More: “Advertisers Say OpenAI Is Extending Its Ad Pilot Beyond April” (March 26) – Adweek

From tipsheet: The run rate roughly equals around $10 million in revenue in 6 weeks. So, it’s still very early.

With the current state of OpenAI’s ad infrastructure (which will only get better), let’s say they open it up to 100% (from 20%) of the US free-tier users. I’ll guess revenue is $500 million annualized.

And then open ads to the world – I’ll guess that revenue triples to $1.5 billion annualized given the number of free tier users. Still very very early.

CREATIVE

Meta adds AI to its creative toolkit

At IAB’s NewFronts, Meta made AI the star of the show according to a blog post coinciding with its afternoon presentation yesterday in New York City.

Creators

First, partnerships tool updates for creators and brands were highlighted including a new tentpole event framework reminiscent of traditional TV.

Meta is looking to not only grow its creator base and the corresponding ads/sponsorship revenue, but also provide an attractive alternative to creators on Google’s YouTube (whose NewFronts’ presentation followed Meta’s).

More creative is better

Next, for brands and agencies, the company emphasized that more creative is better for AI and its Advantage+ ad system.

Meta has added “advanced generative AI video and translation features, including video voiceover and UGC-style videos with avatars, to help brands scale video production and give our AI more to work with to maximize results.”

Translation for localized advertising is made easy courtesy of Meta AI: “A new streamlined flow in Ads Manager allows advertisers to translate all elements of their creative in a single step, making it easier for businesses to connect with people in their preferred language.”

The latest creative updates mean no less than 17 different AI creative formats are available within Meta’s AI ad platform Advantage+ including animation, music, overlays and call-to-action (CTA) enhancements.

See them all: “About Advantage+ creative” – Meta Business Help Center

Read the blog post: “From Commerce to Culture: How AI is Powering Personalized Connection and Growth Across the Funnel” (March 26) – Meta

Meta said yesterday’s NewFronts presentation built upon its announcement at ShopTalk earlier in the week. Read more from tipsheet.

From tipsheet: On the Creator announcements, one can’t help but think that it’s still a manual process in bringing together creators and brands at Meta even with the new partnership hub. AI to the rescue at some point?

A startup called Agentio is pitching itself as an “AI-Native Platform for Creator Advertising” and trying to bridge both YouTube and Meta’s Instagram via a creator/brand partnership hub run by AI agents.

Could Agentio be a nice tuck-in for Meta? Will Meta build or buy its agentic infrastructure for creators?

Manus, an AI agent startup acquired by Meta in December would seem to be well-equipped for the agentic task of connecting brands and creators assuming it can overcome its current challenges with China.

On the other hand, if Agentio can secure the favor of loyal, differentiated creators and brands and drive some decent revenue using agentic workflow, perhaps it would be attractive to Meta or Google, for that matter.

TECH

Adora cranks creative with AI

In a story at the crossroads of AI and creative personalization, AdExchanger covered hotelier Wyndham and lifestyle brand goop using startup Adora, which touts itself as an AI-enabled “performance marketing engine.”

Goop CMO Alexa Ritacco told AdExchanger’s Joanna Gerber:

“The ‘sheer number of iterations’ that the brand needed to test was becoming ‘operationally heavy,’ she added, and it was detracting from the amount of time the team was able to spend on more complex, higher-level creative projects.

Now, they’re able to offload the more “manual, repetitive work” and put their energy into developing thoughtful creative, she said.”

This is all about using AI to mass produce creative which is guided by Adora’s performance engine as it searches for the perfect message for the consumer.

The plug-and-play nature of Adora seems like a potential differentiator. For example, the company’s platform taps the advertiser’s preferred measurement system as Gerber explained:

“According to Andrew Double, VP of revenue at Adora, the platform syncs with an advertiser’s existing measurement systems (often through Adobe and/or Meta) to provide running feedback throughout a campaign. It also defers to the advertiser’s KPIs.”

Read: Wyndham And Goop Are Using AI To Scale Creative And Avoid Burnout (March 26) – AdExchanger

More: Wyndham Taps Adora to Power Next-Gen Marketing (March) – press release

From tipsheet: From here, the steady training of AI model(s) across clients would seem to further refine its performance marketing engine product. A fine wine takes time.

PLATFORMS

Cognitiv targets audience, context with AI

AI ad tech firm Cognitiv, which provides a predictive targeting engine and demand-side platform (DSP), announced new integrations with Index Exchange, Magnite and LiveRamp for Cognitiv’s AudienceGPT product.

Cognitiv unpacked the details on LinkedIn:

“Marketers waste millions targeting who consumers were, rather than where they are in their journey today. That changes now. We are launching AudienceGPT, an AI-powered tool that predicts consumer journeys to create dynamic audiences, so targeting stays aligned with where consumers actually are.

Describe your target audience in plain language and activate across CTV, audio, social, and programmatic. It really is that simple.”

Read more. (March 26)

Audience vs. Context

A press release discussed the difference between AudienceGPT and ContextGPT, the complementary side of Cognitiv’s predictive engine: “While ContextGPT determines the most relevant moments within content, AudienceGPT predicts where consumers are in the purchase funnel and what products they may be interested in.”

Read: “Cognitiv Declares the End of Static Audience Segments with the Launch of AudienceGPT” (March 26) – press release

From tipsheet #1: The whole “audience” targeting discussion has been sidelined given the rise of OpenAI’s ChatGPT and what’s thought to be possible with advertising and LLMs.

AI powered by LLMs doesn’t know you like a cookie or email match does. It predicts based on patterns. But why not use both —context and audience— in a privacy-safe way to improve addressability? This appears to be Cognitiv’s point.

From tipsheet #2: Context and audience will converge inside ChatGPT someday.

MEASUREMENT

AI powering performance in TV

In The Trade Desk’s publication The Current, Travis Clark parses NewFront presentations from Samsung, LG Ad Solutions and Comcast and sees an “AI-powered performance TV” pattern.

After explaining his premise, Clark observed:

“To that end, Samsung announced a ‘full-funnel performance platform powered by AI and shoppable innovation’; LG Ad Solutions unveiled Own the Outcome, a CTV performance framework; and Comcast Advertising launched Outcomes+, (…) a suite of AI-powered products.

Read: AI-powered performance TV takes over NewFronts talking points (March 26) – The Current

Clark’s point about outcomes and performance is well-taken.

Late Wednesday, Amazon Ads VP Kelly MacLean, who oversees engineering, product, and “science” at Amazon DSP, discussed her NewFronts appearance with Samsung on LinkedIn:

“At IAB NewFronts, I joined Courtney Howell from Samsung Ads to share that starting in July, Amazon Ads’ interactive video ad capabilities will be available on Samsung TV Plus — the first time we’re extending these experiences beyond Amazon-owned properties.

Viewers can engage with brands and take action directly from their Samsung TV, whether that’s adding a product to cart, sending information to their phone, or signing up to learn more. For advertisers, this creates a clearer link between engagement and outcomes…”

Read more on LinkedIn. (March 25)

Related: “Amazon is Looking to Leapfrog Walmart in Shoppable TV” (March 26) – Mike Shields in his “Next in Media” SubStack

DATA

Opinion: Enterprise data makes the AI difference

Snowflake, with its AI data cloud capabilities and consumption-based pricing, powers much of the advertising and marketing technology ecosystem today.

In an Adweek op-ed, Snowflake Industry Principal Mike McCarver may be talking Snowflake’s ‘book’, but he reminds enterprise readers that incorporating enterprise data is critical in working with AI.

First-party data — the enterprise’s proprietary data — is the differentiator.

Mr. McCarver wrote:

“Without enterprise data, your AI models are commodities

That context often comes with access to enterprise data. Whether it’s structured or unstructured, that data—including historical performance data, customer behavior, or business constraints—can provide much more specific outputs around your needs.

Ask an AI coding assistant without access to your data to build you a bespoke analytics application, and it’ll give you something polished, possibly even technically sound, but it won’t reflect your reality….

Read: Opinion – “Your AI Model Is a Commodity, But Your Data Is a Differentiator” (March 26) – Adweek

TECH

Seufert: Defining ‘agentic commerce’

As retailers climb aboard the agentic ‘freight train,’ the definition of agentic commerce appears to be broadening beyond agents executing transactions. Mobile Dev Memo analyst Eric Seufert has had enough.

A March 24 announcement by payments company Stripe about a new Facebook checkout partnership pushed him over the edge.

Seufert railed on X yesterday:

“If Instant Checkout didn’t represent agentic commerce, using that term to describe Meta’s new ‘purchase-from-an-ad’ experience is simply absurd. An agent plays no role in this transaction: a user sees an ad and clicks a button to make a purchase. This workflow merely resurrects the social commerce model Meta had previously abandoned, with no agent involved whatsoever. The fact that this interaction format is being classified as agentic commerce underscores how diluted, distorted, and devoid of analytical rigor that phrase has become.”

Read it on X. (March 26)

More: Meta partners with Stripe for in-ad shopping (March 26) – Mobile Dev Memo

From tipsheet: Agents mean there should be autonomy.

The fluidity of industry definitions speaks to the fast pace of AI transformation. Agentic advertising will, no doubt, run into this issue.

“Bring on the fluidity!,” said the trade publisher.

PEOPLE MOVES

Now hiring

MORE

  • TikTok rebrands its advertiser pitch around full-funnel ambition (March 26) – Digiday (subscription)
  • OpenAI Killed Sora. Creatives Had Already Moved On (March 26) – Adweek
  • From media quality to revenue optimization: IAS and Mastercard are turning media signals into real sales performance (March 26) – Integral Ad Science
  • “Today we are open-sourcing our Agentic Intent Classifier (AIC) model. This is something we built at AdMesh to understand intent inside AI conversations in real time, and we believe it should be open for everyone to use.” (March 25) – Manikumar Gouni, CEO, AdMesh on LinkedIn