Commercial AI Briefing: September 2026

Executive read

ChatGPT advertising expanded across Southeast Asia and Taiwan. OpenAI began testing Sponsored Agents and added AI support inside Ads Manager. Google opened AI Max for Shopping as a beta, announced richer AI Max reporting, and continued selected Search campaign upgrades. OpenAI also released a public beta for the Agents API and introduced GPT-6 Sol and Luna.

Each development gives a business more latitude: a new place to reach buyers, a new way to generate or route an advertisement, a new agent runtime, or a lower-cost model tier. Leadership still has to define the customer, approved action, public claim, conversion value, and evidence that support the investment.

Conversational advertising arrived in Southeast Asia

OpenAI announced on 23 September that ChatGPT advertising would begin rolling out in Singapore, Indonesia, Malaysia, the Philippines, Thailand, Vietnam, and Taiwan. The company said self-service access was available to eligible businesses, with advertisements shown on Free and Go plans while Plus, Pro, and Enterprise remained advertisement-free.

OpenAI says advertisements are labelled, separated from answers, and subject to privacy controls. Advertisers remain responsible for inspecting the live placement, market eligibility, landing-page journey, customer response, and commercial result. Product availability, fulfilment, language, pricing, and service promises require market-level accuracy before conversational demand reaches a seller.

Advertising operations gained new agents and assistants

OpenAI announced on 16 September that it was testing Sponsored Agents with selected United States advertisers. It also described an Ads Manager plugin, AI creative suggestions, opt-in text customisation, and a HubSpot connection. The company said its Shopify application would begin international rollout where ChatGPT advertising was available from 23 September.

An agent that can help build or modify a campaign changes the approval path. The business requires a clear distinction between analysis, recommendation, and an action that changes spend, audience, creative, or destination. Generated text carries the same claim-proof and landing-page requirements as copy written by a person.

Google widened AI-assisted shopping

Google opened AI Max for Shopping as a beta in September. Google's help documentation says it can customise product titles, expand final URLs to other commercial pages, and serve additional text formats. It also provides brand exclusions, URL exclusions, term exclusions, messaging restrictions, and retail-focused reports for generated titles, search terms, landing pages, and expanded assets.

Text customisation is currently limited to English feeds, while Final URL Expansion requires Target ROAS and text customisation. AI Max for Shopping is a beta suite for Shopping campaigns; AI Max for Search operates inside Search campaigns. Any result comparison needs the exact campaign type, enabled features, feed, bidding strategy, tracking template, and destination history.

AI Max reporting and migrations required dated account records

Google announced on 23 September that a unified AI Max report would connect search terms with generated creative and selected landing pages. Google said availability details would follow later in 2026. It also expanded the AI Brief closed beta to seven additional languages.

Google continued automatic September upgrades for eligible campaigns using standalone automatically created assets and campaign-level broad match. The Dynamic Search Ads sunset and automatic upgrade were deferred to February 2027. Marketing teams need a dated configuration record for Search Term Matching, Text Customisation, Final URL Expansion, brand controls, exclusions, and the migration date before comparing periods.

Agent development and model economics changed together

OpenAI released the Agents API in public beta on 10 September. The company described a managed coding-agent harness, sandbox options, long-running sessions, tool search, programmatic tool calling, and multi-agent support. On 22 September, OpenAI introduced GPT-6 Sol and Luna, with API prices described as 50 per cent lower than GPT-5.6 promotional pricing.

An agent that can search, call tools, and run for a long session requires a precise authority. Cost comparisons also require the actual context, tool use, retries, review time, and error consequence.

How do AI Max and Performance Max differ?

Google describes AI Max as an optimisation layer for existing Search campaigns. It combines broader search-term matching, text customisation, and Final URL Expansion with brand, location, URL, and reporting controls. Google says Final URL Expansion can send a user to the page on the advertiser's domain that the system predicts will perform best.

Microsoft made its AI Max suite generally available in August 2026. Microsoft says its version also combines expanded search-term matching, text customisation, and Final URL Expansion inside Search campaigns, supported by reporting, brand controls, term exclusions, and existing campaign settings.

More routing freedom gives the platforms more opportunities to expose unclear pages, inconsistent claims, weak exclusions, and conversion goals that reward low-value activity. The website, offer, and customer-value definition set the direction.

Where does Performance Max operate?

AI Max enhances Search campaigns. Performance Max is a separate goal-based campaign type that works across eligible inventory.

Google lists YouTube, Display, Search, Discover, Gmail, and Maps among its Performance Max inventory. Microsoft describes Performance Max as a cross-channel campaign across its advertising inventory, including search, native, display, audience, shopping, and other eligible surfaces.

This distinction affects budget and diagnosis. Search-focused expansion can be examined through search terms, headlines, landing pages, and query intent. Cross-channel optimisation also requires evidence about where spend moved, which placements and assets contributed, and whether the resulting customers justified the broader reach.

Which conversion does the platform learn from?

Google's enhanced conversions for leads can match hashed first-party lead data with imported offline outcomes. This allows an advertiser to connect an ad interaction with a later customer event, such as a completed sale, when the implementation and privacy basis are sound.

Microsoft's Universal Event Tracking supports conversion goals, audience creation, and automated bidding. Microsoft advises Performance Max advertisers to choose meaningful conversion goals and supply enough reliable conversion data for the campaign to learn.

The practical task is to decide which event most closely represents commercial value. A submitted form, booked meeting, qualified opportunity, completed order, renewed contract, and profitable customer describe different stages of the journey. The platform can optimise only against the evidence it receives.

Where does platform measurement stop?

On 10 September 2026, Google introduced a Data Strength Uplift metric that calculates additional conversions recovered through an advertiser's first-party data setup. Google presents causal experiments and marketing-mix modelling as separate tools for evaluating incremental impact.

That boundary matters. Better attribution can recover conversions that were previously difficult to connect to advertising. Commercial quality still requires separate evidence about margin, sales qualification, retention, and incrementality.

Management has to connect platform reporting with CRM progression, revenue, cost to serve, and customer value before increasing spend.

What should commercial, marketing, and leadership teams inspect?

Commercial and sales

Define the customer and outcome that justify acquisition spend. Return reliable downstream events to the measurement process where the legal, privacy, and technical basis is in place.

Marketing

Review the searches, generated messages, selected landing pages, exclusions, channel allocation, and conversion goals that guide each campaign. Similar product names coexist with different inventory, settings, reporting, and performance across the two platforms, so each imported campaign requires a fresh review.

Leadership

Set the investment rule before automation expands. The rule should state which customers count, how value is calculated, which evidence supports additional budget, and who can change the goal the platform is pursuing.