Check the market before planning the campaign
Google announced on 28 October 2026 that it was expanding AI Mode ads to Australia, Canada, India, Kenya, Malaysia, New Zealand, Nigeria, Pakistan, the Philippines, and Singapore. Brazil, Mexico, and the UK were listed as planned Q4 additions. It also announced ads in follow-up conversations in the US. Check your account before treating a planned launch as available inventory.
Google says AI Mode questions averaged three times the length of traditional searches in its May 2026 US research. Longer questions can tell you more about the need, but they don’t automatically make a better sale. I’d start with the customer, market, and offer you can actually reach.
Understand what you’re buying in Google
Google’s documentation makes relevant Search, Shopping, and Performance Max campaigns eligible for these placements. You can’t exclusively target, or separately opt out of, AI Overviews and AI Mode. Their impressions, clicks, conversions, and costs feed into standard campaign reporting together with other activity.
Some sponsored options also have independently generated AI explainers that the advertiser can’t manually edit or approve. Google says buying ads doesn’t influence its organic AI answers. So I’d distinguish the placement you buy from the explanation the system produces, and be careful about crediting an aggregate campaign improvement to AI placements alone.
Check advertiser access and audience access separately
OpenAI’s Ads Manager guidance bases self-service eligibility on the country of the legal entity advertising and being billed. It lists eligible countries and separates account creation from getting access. A successful signup doesn’t confirm every audience, format, or category is open to your business.
I’d check three things before spending time on creative: the billing entity can advertise, the target market is available, and the specific offer is eligible. If your company operates internationally, write those answers down for the market you actually want to test.
Keep paid placement and the AI answer separate
OpenAI says ads run separately from the chat model and can’t shape or rank its answers. Its current FAQ describes ads for eligible Free and Go users, with the listed paid individual and business subscriptions remaining ad-free. It also describes testing a separate multi-image placement during image generation.
Advertisers get aggregated performance information, rather than people’s chats, histories, or memories. A relevant placement gives you a chance to be noticed. You still need the person to choose a next step. I wouldn’t build a sales plan around the idea that an ad gives you access to the conversation or an AI endorsement.
Read the outcome behind each case-study number
OpenAI’s 5 October announcement reports partner findings for WeightWatchers, Dose, and Portland Leather. The table below keeps each result attached to what it measured. A lower attributed acquisition cost, the customer mix within additional purchases, and the share of new visitors are all useful questions, but they give you different answers. These are platform-published findings, not an independently audited advertiser average.
| Business and reporting partner | Published finding | What the number measures |
|---|---|---|
| 15.3% lower attributed acquisition cost than its blended paid-search benchmark | Cost assigned to an acquisition under the measurement method | |
| 67% of incremental purchases came from net-new customers | The customer mix within purchases estimated to be additional | |
| 93% of visitors from ChatGPT Ads were new | Visitor newness, not a purchase or profit rate |
Use your margins to decide what the result is worth
A new visitor may never buy. An attributed customer may have bought anyway. And a new sale may still cost too much to fulfil. Before the test, I’d agree what counts as success after advertising, delivery, and service costs, using the event your business can actually measure.
For a longer sales cycle, you may need to review suitable inquiries first and completed sales later. Keep the connection clear rather than treating every form submission as revenue. The published cases show useful outcomes are possible; your offer and economics decide whether the same channel is worth your investment.
Give the measurement window some thought
OpenAI’s measurement article says 52.7% of eligible one-day view-through conversions in an early analysis happened within an hour of the matched impression. That counts a group of attributed conversions; it doesn’t say 52.7% of ad viewers bought. The article also reports a WorkMagic geographic-lift study estimating 2.3 times more additional Dose orders than last-click tracking captured.
One method connects an impression with a later event inside a window; the other estimates extra business through an experiment. I’d discuss which method you’re using before the test starts. A purchase soon after an impression is useful evidence, but timing alone doesn’t prove the ad caused it. OpenAI describes this work as early-stage, so don’t apply its multiplier to your own sales.
Check the policy for your actual offer
OpenAI’s policies, updated on 10 September 2026, restrict health services outside the US and describe case-by-case approval for some US health advertisers. They also prohibit individual housing rental and sales ads. Separate rules exclude vulnerable contexts, including personal and mental health conversations.
For senior living, I’d get the platform’s classification of the exact offer and market before building a budget around it. A retail case doesn’t settle whether a care-related campaign is eligible. Where paid placement isn’t available, families may still need clear public information and accurate organic answers.
Ask what the campaign added
A detailed conversation can help an ad meet a specific need. It can also put the ad in front of someone who was already close to buying. Both can look good in an attribution report, even though they may add different amounts of business.
Picture two fictional buyers. One was already planning to purchase and sees an ad on the way. The other discovers a suitable offer through the ad and makes a purchase they otherwise wouldn’t have made. Both purchases could receive advertising credit; only the second adds a sale in this simplified example. Keep that question in the review, especially when existing brand demand is strong.
Here’s how I’d make the decision
Confirm access and policy fit, choose a specific offer and audience, and agree the result you’ll judge. Keep the budget limited enough to learn from. Review the available reporting, inquiry or customer quality, and contribution after costs. Expand because the evidence supports it, rather than because the platform is new.
I work at Searchable and can help connect your paid test with the wider questions buyers ask about you. Accurate information, useful organic discovery, and paid exposure each have a role. If you need to put the whole strategy together, return to the implementation guide and start with your buyer’s actual decisions.
Keep availability and results in context
The sources were reviewed on 3 November 2026. Availability comes from official documentation and dated announcements; planned launches remain plans. Partner results retain their original measures and haven’t been independently audited here. The fictional buyers illustrate attribution, rather than describe real customers. Recheck access and policy in your account before running a campaign.
Sources and details
Sources reviewed: November 3, 2026. I work at Searchable. The examples and recommendations are mine; reported results are credited to their sources.
1
Searchable