August 13, 2026

Microsoft plans to increase production of Maia 300, its next-generation inference-focused AI chip, ahead of a fall release, The Information reported, citing an unnamed source. Maia 300 follows Maia 200, announced in January as part of Microsoft's "heterogenous AI infrastructure" for running AI models including OpenAI's latest systems. Training builds an AI model. Inference runs it, generating the predictions and answers users see in products like Copilot and search. The hardware divisions at Microsoft, Google and Meta describe the current period as "a major economic and technical turning point," since the cost and computing power needed to run AI models has overtaken the investment needed to train them, a shift known as the "inference flip." Meta has built its own inference chip, Iris, reported to handle content ranking, ads and generative AI across Facebook, Instagram and WhatsApp. Custom silicon aims to cut the cost of running AI at scale.

AppsFlyer now recognizes Tatari as both an Ad Network partner and a Signal Partner, a dual status the company says no other CTV platform holds. The two classifications determine how TV campaigns show up inside AppsFlyer, the mobile measurement platform most app marketers use to justify spend to leadership. As an Ad Network partner, Tatari's TV impressions carry an AppsFlyer tracking URL and AppsFlyer runs its own attribution, a model built for programmatic streaming that leaves linear TV and direct streaming buys outside the data. As a Signal Partner, Tatari instead receives installs and conversion data from AppsFlyer and runs its own incrementality-based measurement on top, adding a proprietary device graph and IP filtering designed for TV. Marketers can use either model, or run both side by side to compare the numbers.

Nielsen has switched on streaming ad measurement inside Digital Ad Intel in Brazil, following earlier launches in the US and Europe. Sabrina Balhes, Managing Director of Nielsen Brazil, said the company holds the only reference-standard position for streaming ad tracking in the country. In the first three months of monitoring, Nielsen logged more than 23,000 creatives across nine streaming apps. The new coverage gives brands visibility into share of voice, communication intensity and competitive moves inside streaming environments, data advertisers can use to guide spending toward market share gains, expansion or scale efficiencies. Balhes said the volume of tracked creative "really speaks to the robustness" of Nielsen's coverage. Brazilian advertisers now have a new benchmark for streaming competitive intelligence in the country.

SE Ranking studied 50,006 commercial prompts across 20 U.S. niche markets and found that 14.35% of ChatGPT ads were topically unrelated to the query that triggered them, with the mismatch rate climbing above 50% in categories like Relationships and News/Politics. Advertisers appeared as a cited source in ChatGPT's answers in just 3.63% of ad placements, compared with 11.53% in Google AI Mode. The study identified 1,159 unique advertisers testing the format, led by fintech platform BestMoney, which accounted for 13.56% of all ChatGPT ad units. SE Ranking also ran its own two-week test, three campaigns, eight ad groups and 48 ads across the U.S., Canada, Australia and New Zealand, generating more than 97,000 impressions and 1,263 clicks but very few sign-ups. ChatGPT ads currently reach only Free and Go plan users.

Castle Walls, a two-episode historical drama produced by Turkish company Ay Yapım, premieres on Prime Video on Wednesday, billed as the first show made end-to-end with artificial intelligence to launch on a major global streaming platform. The series, titled Surlar in Turkish, is set around Alamut Castle in present-day Iran, a location also featured in Ubisoft's Assassin's Creed Mirage. It was written by Kerem Deren and Şisil Hazal Tenim and produced by AI Yapım, Ay Yapım's Singapore-based AI subsidiary. Production used a proprietary storyboard system and AYDNA, an in-house video diffusion model trained only on Ay Yapım's own licensed content. Ay Yapım says every element of the training pipeline used rights-managed material, and human actors performed all roles in the series. Human voice artists handle dubbing in every language.

Time began selling ads inside the markdown version of its webpages, the format AI agents read when they crawl a site, with Ally Bank and the Project Management Institute as early buyers. Perplexity confirmed it blocks those markdown ads from influencing its search index and AI agents. Chief communications officer Jesse Dwyer said Perplexity works to protect users from deceptive practices, sponsored or not, and that publishers deploying markdown ads risk a reputational downgrade, including a hit to their trust score. Neither company detailed how the blocking works or whether it changes Time's existing ad deals. The ads, built by ad tech firm Mobian, insert brand-written FAQ content into Time's markdown pages so AI agents read it as fact alongside the publisher's own reporting.

Eighty per cent of marketers say they cannot fully explain why their campaigns succeeded or failed, according to a survey of more than 250 advertising professionals by Brand Lift measurement company On Device. Sixty-three per cent rely on educated guesses, such as sales spikes or click rates, to interpret results, and 16 per cent move straight to the next campaign without learning why the last one worked. Marketers already running Brand Lift studies reported far higher confidence: 28 per cent said they were completely confident explaining outcomes, compared with 14 per cent of those without Brand Lift measurement. Among marketers without Brand Lift, 28 per cent described themselves as flying blind after campaign sign-off, against 2 per cent of Brand Lift users. On Device CEO Alistair Hill said the results point to gaps in what the industry measures.

Public comment on the changes stays open through September 11, 2026. IAB Tech Lab released updates to its Privacy Standards Portfolio, revising the Global Privacy Protocol (GPP) and finalizing version 2.0 of the Data Deletion Request Framework (DDRF). The GPP revisions align with the Fifth Amended and Restated Multi-State Privacy Agreement, removing the prior state-by-state coverage model, eliminating Service Provider and Opt-Out Option Modes, dropping secondary usage consents, and simplifying notice and choice fields. DDRF 2.0 follows a comment period that opened in fall 2025 and adds clearer identity and deletion request JWT definitions, improved result feedback and troubleshooting, stronger framework integrity, and support for implementation-specific extensions. CEO Anthony Katsur said the changes make compliance "more consistent, transparent, and practical across the ecosystem." The Global Privacy Working Group and the Privacy Rearc Commit Group developed the updates.

India's out-of-home advertising market is losing advertiser confidence over measurement and verification, according to a report published by invidis.com. The report identifies five recurring problems: media owners inflating traffic figures for billboard locations, installation delays and early removal of creative before contracts end, the same structure sold to multiple clients at once, campaigns placed at inferior or substituted locations, and proof-of-posting images manipulated or recycled with AI editing tools. Times OOH and JCDecaux are cited as operators with strong reputations for professionalism, while mall signage specialist Skycom is named as one of the few companies using Quividi for network-wide audience measurement and proof of play. The report states that 'the biggest challenge for India's OoH industry is not technology, inventory, or demand. It is trust.' Independent verification and standardized reporting are described as prerequisites for the market's long-term credibility.

Yahoo DSP and Clinch expanded their partnership on Wednesday, letting advertisers target two audience segments from a single campaign by uploading one set of creative assets and letting the platform choose the right ad at the moment it serves. Yahoo DSP now sends audience data to Clinch in real time through the platforms' two-way API, and Clinch uses that data to select the correct ad for each impression. Dave Modica, Yahoo's director of business development, said advertisers previously needed separate campaign lines and separate creative to reach groups such as new moms and sports fans. First-party, CRM, lookalike, modeled and third-party audiences housed in Yahoo DSP can now drive that decision at the millisecond an ad renders. Campaigns and line items sync automatically between the two platforms, and creative tags publish across display, video and audio without manual trafficking. Clinch joined the Yahoo DSP Agent Network, which links 23 technology partners, in June 2026.