Google AI Mode Can Wait for a DealAn alert arrived. Is it actually cheaper?

01Search gets a second verb: wait

Your kid’s sneakers are sold out in the right size. The headphones still cost too much. One search cannot finish either job. Google Search now takes a second kind of request: leave your conditions and hear back when something changes. Searching once becomes watching over time.

Price tracking and alerts are not new: Google described them in November 2025. The news is who gets AI Mode’s ongoing info monitoring, not the invention of the price-drop alert.^[1]

On September 28, 2026, Google Search VP Robby Stein said AI Mode’s info monitoring, previously used by Pro and Ultra subscribers, was rolling out to everyone globally. Tell it what to look for, and Search keeps checking the web and lets you know when something relevant changes. His examples: new restaurants or pop-ups nearby, local holiday activities for the kids, and back-in-stock and price-drop updates.^[2]

At I/O in May, Google described information agents operating in the background, 24/7. That means you do not have to keep the app open. It does not mean every website is checked every second. We found no published commitment on how quickly an alert arrives.^[3]

A shopping radar cuts the time you spend watching. It cannot tell you whether the item is worth your money right now. This article focuses on the gap between the blip on the radar and the payment.

Extra: what does the 60-billion figure count?

Google’s May 2026 Universal Cart announcement describes a Shopping Graph of over 60 billion product listings. Listings are not unique models or stores. One product can appear in multiple offers and configurations. Google also uses this catalog for Universal Cart’s price-drop and back-in-stock experiences.^[4]

The often-repeated “2 billion updated every hour” figure comes from a November 2025 announcement, when Google described the total as over 50 billion. Pairing that older refresh rate with the newer 60 billion figure would suggest a current, verified update speed that Google has not published.^[1]

02Keep watching, check now, buy: three different requests

Waiting for a price drop, checking today’s stock and placing an order are different jobs. Info monitoring tracks changes over time. Check for me gathers current information. Agentic checkout executes a purchase after confirmation. Opening the door wider for one does not relax the conditions on the others.

Use the buttons below to turn the same headphone request into three different ones. This is an editorial demonstration, not Google’s interface or a connected agent.

“Check for me” gathers information from the web and can call a business when online details are missing. It is not a calls-only feature. The help page says it is available in English to signed-in users over 18. Web results can take a few minutes; call results can take up to 30 minutes. Add “don’t call” if you want online checking only. Those conditions and timings describe this check-now feature; they are not a promise about ongoing monitoring.^[5]

The help page recommends turning on history and personalized recommendations “for the best experience.” Before approving a request, review what will be checked and whether calls may be made.

03Is the first price AI shows really the cheapest?

21.6%. Read on its own, it sounds like a surcharge for letting AI shop for you. That is not what was measured. Productrise, a company that tracks product visibility in Google, compared product prices displayed in AI Mode and regular Google search, not monitoring alerts. From August 9 to 31, 2026, it collected more than 100,000 search results and AI Mode responses in the US and UK, covering over 2 million product listings.^[6]

For pairs matched by query, date and Google product ID, AI Mode’s first displayed offer price, or lead price, averaged 21.6% higher. But a matching product ID does not guarantee matching purchase terms. Some pairs compared a new item with a used one.^[6]

So this is neither money shoppers actually overpaid nor the premium for buying the identical offer. It is a reason to look beyond the first displayed price, not a tax on using AI.

In 49.6% of matched pairs, the first-listed seller was different too. Finding the same product is not the same as showing the same seller’s terms. The number of choices differed sharply as well: an average of 3.9 product cards per query in AI Mode versus 27.8 in regular search’s popular-products carousel. A shorter answer is convenient. It does not certify that the offers you never saw were worse.^[6]

Study notes: the comparison has boundaries

• Regular-search collection covered the popular_products carousel; AI Mode collection used all tracked product cards. It did not cover every shopping placement.
• The introduction says both sides were collected at the same moment, while the methodology specifies matching on the same date. The public methodology does not let us verify identical capture times for every price.
• A shared product ID does not guarantee the same seller, condition or purchase terms. The source includes new-versus-used examples.
• The 21.6% result covers matched products. The 49% full-catalog median gap is a different comparison. The source prints medians of 149 and 100 but says USD and GBP were not converted. We do not apply that mixed-currency statistic to a shopper’s dollar budget.
• The 38.1% disagreement figure does not neatly reconcile with another sentence saying prices agree “about 68%” of the time. The source does not explain the difference.
• This is one US/UK observation window. It cannot tell you what a search in Canada or Korea shows today, or what holds for every product category.

Source: Productrise’s original study and methodology. We did not independently reproduce it.^[6]

04Does the cheapest price tag make the cheapest receipt?

Seller A has the lower sticker price. Add shipping, subtract discounts, and B comes out cheaper. The cheapest price tag does not always make the cheapest receipt. First match the model, condition, warranty and return terms, then compare what you would actually pay delivered to your address.

Try changing the prices, shipping and discounts in this fictional USD example. Watch the winner change. You can also enter an assumed tax rate. This is not a recreation of Productrise’s 21.6% result.

This is not about redoing all the work the agent saved you. Let it handle the waiting while you handle the last comparison. Treat an alert as a starting signal, not a certificate of the lowest price.

05So where does checkout finish?

Once the price checks out, one question is left: “So where do I pay?” Sometimes the order finishes inside the AI conversation. Sometimes you are sent to the merchant’s own store. Start with the current paths for Shopify products.

Two checkout paths for Shopify products

Based on Shopify help checked October 4, 2026. This is not a comparison of everything UCP and ACP can do.

In March 2026, OpenAI said the initial version of Instant Checkout did not offer enough flexibility and shifted its focus to discovery and merchants’ own checkout. Some merchants, such as Walmart, go deeper with a dedicated ChatGPT app, so the Shopify path above does not stand in for every ChatGPT purchase.^[9]

Paying inside an AI conversation does not make Google the seller. UCP documentation says the merchant remains the Merchant of Record and keeps customer relationships and data. Who sells the item, and on what terms, matters more than where the checkout window appears.^[10]

Between an alert and payment comes confirmation

Google Pay’s agentic checkout runs only after you confirm the purchase and shipping details. Its help page says that if the final price rises above the total you confirmed, the purchase does not go through and you are not charged. Info monitoring, which sends alerts, and “Buy for me,” which makes the purchase, are different features. An alert is not purchase authorization.^[12][1]

06For sellers: fix the price tags before polishing the copy

Questions after the alert

Can shoppers in Canada or Korea use the same features?

Google announced a global monitoring rollout, but we could not confirm that it has reached every language and account yet. Checkout has separate eligibility. Shopify’s Google direct checkout requires US stores and US customers; other customers go to the merchant’s website. Check your actual account and retailer support.

How should I align dollar prices and delivery locations?

Distinguish USD from CAD and compare the same address, item condition and seller terms. Include shipping, tax, duties and card conversion fees where relevant. This article’s calculator is a fictional USD example, not an exchange-rate or duty calculator. It is not the same budget as CAD 200 in the sample request.

Should I keep watching an item after buying it?

If you no longer need alerts, clean up any remaining monitoring task. If the retailer offers price protection (a refund of the difference if the price drops within a set period after purchase), decide whether to keep watching until that window closes. Find stop or delete controls in the actual task interface and help.

What if the alert price differs from the checkout total?

Match the item, condition, seller and destination again, and stop if the final total exceeds your budget. Google Pay’s agentic checkout says a final price above the confirmed total stops the purchase with no charge. That is not a universal rule for every AI shopping flow.

Final quiz: what comes after the alert?