Snap moves into DMs, Amazon opens ChatGPT ads: Sept. 11

Paid media is moving closer to the conversation that precedes a purchase. Snapchat can now place a retailer's catalog inside Chat, while Amazon is opening a managed route into ChatGPT ads. Google, meanwhile, is testing a safer way to change budgets and bid targets. The upside is shorter buying paths. The bill arrives when attribution gives every platform credit.

Snapchat puts the product shelf inside the inbox

Snapchat made Dynamic Product Ads in Sponsored Snaps available to advertisers worldwide on September 10. Dynamic Product Ads pull items from a retailer's catalog and choose which product to show. The company's Commerce Power Pack announcement says those ads can now appear in Chat, a surface used regularly by 85% of Snapchatters.

That moves the product shelf into a place people associate with personal messages. For retailers, fewer steps between seeing an item and opening its page can lift conversion rate, the share of visitors who buy. But the placement carries a sharper creative risk: an irrelevant catalog item in an inbox feels more intrusive than the same ad in a scrolling feed.

Snap also put Value Optimization into beta for web and app campaigns. It tells the system to pursue higher-value purchases rather than the largest count of purchases. Useful, if the purchase values sent back to Snap are accurate. Returns, discounts, and repeat buyers can make reported return on ad spend look richer than the cash the campaign actually created. That is the same measurement gap covered in our small-business attribution framework.

Amazon gives advertisers a managed door into ChatGPT

Amazon's demand-side platform will carry ChatGPT ad inventory for a limited group of U.S. advertisers, Marketing Dive reported on September 10. A demand-side platform is software used to buy ads across outside publishers. Amazon will manage campaign setup, while OpenAI will decide where the ads appear inside ChatGPT.

The pilot accepts both cost-per-click buying and CPM buying, meaning what an advertiser pays to show an ad one thousand times. It also includes catalog-fed product ads. That creates a new distribution path for businesses already using Amazon's ad-buying system, but it splits control: Amazon handles the buy, OpenAI handles delivery, and the advertiser must still decide whether a sale was incremental.

The reporting described for the pilot includes impressions, clicks, costs, CPM, and cost per result. Those are operating signals, not proof of profit. Businesses testing this route should compare closed sales and gross profit against other paid channels, then count each buyer once. Our earlier brief on ChatGPT's European ad expansion explains why a new auction deserves its own source field before it deserves more budget.

Google lets Search advertisers test the money dial

Google Ads is testing experiments that compare different budgets, target CPA levels, or target ROAS levels across Search campaigns, Search Engine Roundtable reported on September 10. Target CPA is the amount you tell Google you want to pay for one sale or lead. Target ROAS is the sales value you ask it to produce for each ad dollar.

This is a better control than changing a live campaign and guessing what caused the result. A structured split test can show whether extra budget finds profitable demand or simply buys costlier clicks. The catch is patience. A sales cycle that takes three weeks cannot be judged from three days of platform results.

The experiment is a beta, so availability is not universal. Accounts that have it should test one economic decision at a time and hold the offer, landing page, and conversion definition steady. Accounts without it lose nothing by waiting. The core controls in our Google Ads budget audit still catch waste before a larger budget amplifies it.

What to watch

• Watch whether Snapchat's inbox inventory keeps conversion quality when holiday ad volume rises. Cheap orders with heavy returns will expose the weakness quickly.
• Watch Amazon's pilot for self-service access and clearer sales reporting. Managed entry can add cost before it proves a lower customer acquisition cost.
• Watch Google's experiment rollout for smaller Search accounts. The value is clean evidence, not another forecast button.