Strategy

AI assistants as an acquisition surface, and what paid teams should measure

AI answers now show ads, send some referrals and are starting to take purchases. Here is what is documented, what is not, and a short list of things to track.

4 min read

People increasingly ask an AI assistant or a Google AI answer what to buy or which app to try, and some of those answers now carry ads, links or checkout. The channel is real enough to track and too immature to budget around. Here is what the primary sources say, followed by a measurement plan that costs little and works even if the channel stays small.

What is documented

Ads in Google's AI answers. At Google Marketing Live on May 21, Google said it was expanding ads in AI Overviews to desktop and bringing ads to AI Mode. Google's help documentation on ads and AI Overviews, as reproduced by Search Engine Roundtable, says text and Shopping ads from existing Search, Shopping and Performance Max campaigns are eligible, that ads must be relevant to the content of the AI Overview, and that you cannot directly target these placements or opt out of them. It also says Google Ads does not offer segmented reporting for ads that show within AI Overviews, and that they are reported as Top Ads. Google recommends broad match or keywordless targeting, smart bidding, strong creative and current Shopping feeds. Check the current help pages for where AI Mode ads stand in your market, since rollouts differ by country and over time.

Referral traffic. Digiday reported in May, using Similarweb data, that referrals from OpenAI's assistant to news and media sites nearly doubled between January and April, reaching 243.8 million visits across 250 sites in April. Publishing executives said the volume was still small compared with other sources. The article reported visit counts only, with no conversion data.

Checkout inside the assistant. On September 29, OpenAI launched Instant Checkout for US users, starting with single-item purchases from US sellers on one marketplace, with more merchants promised. It was built on the Agentic Commerce Protocol, developed with Stripe. Coverage reported that Instant Checkout items are not preferred in product results, and that fee terms were not disclosed.

What is not documented

There is no standard way to see how many of your sales started with an assistant. Google does not segment AI Overviews ad performance. Referral numbers capture only clicks that arrive with a recognizable source, and many people will read an answer and then search your brand name or type your address. In reporting, that visit looks like branded search or direct traffic. Anyone who quotes a precise share of revenue from AI assistants for a typical advertiser is guessing.

A measurement plan that does not require new tools

  1. Segment the referrers you can see. Add the assistant domains to a custom channel grouping in your analytics, and review sessions, conversion rate and revenue per session against your other sources. Expect small numbers and wide error bars.
  2. Ask at the point of conversion. A short, optional "how did you hear about us" field with an option for AI assistants captures what click data misses. It is self-reported, so treat it as directional.
  3. Watch branded search and direct traffic. If more people learn about you from an AI answer and then look you up, those lines move. Track them weekly, and annotate changes in your marketing calendar so you can tie any movement to what you changed.
  4. Keep your feeds and product pages clean. For ecommerce, accurate titles, prices, availability and structured product data serve Shopping ads, AI Overviews and any assistant that reads your catalog. That is a cheap investment that pays off on every surface.
  5. Test, do not infer. If you want to know how much your Google ads inside AI answers add, a geo or time-based holdout on Search or Performance Max is the clean way. Platform reporting will not isolate it for you in the near term.

How much budget to give it

Treat it as a watch-and-learn line, not a channel. A hypothetical rule: if your reporting shows assistant referrals under a few percent of sessions, put effort into tracking and content accuracy rather than a new budget line. If you are an advertiser whose Search campaigns already use broad match or Performance Max, you may be eligible for Google's AI placements already, and you cannot choose to opt out, so your job is to read your Search results with that in mind. A change in click-through or conversion rate on Search could come from a changed results page as much as from your ads.

The practical takeaway: build the referrer segment and the survey question this month, and revisit the budget question when you have a quarter of data.

Sources

More on Strategy

All insights

The Lingviny brief

Platform changes and what to do about them, for people who buy Apple, Meta and Google traffic.