
39% of bookings came from properties that had never converted before
Search Machines · May 2026 · 4 min read
Feed-based advertising exists for a reason. When inventory runs into the hundreds of thousands, keyword campaigns built by hand are not viable. Dynamic Search Ads gave large advertisers the coverage they could not create manually.
For one travel client, that was the setup. Property-level marketing ran through feed-based ads. ROAS was strong. Booking volume was stable.
Then Google confirmed DSA will sunset in September 2026. The replacement is AI Max.
AI Max offers more automation – but less visibility into what’s triggering your ads or where your budget goes
For advertisers who need to know both, that is not a straight swap.
Why the obvious answer didn’t work
The standard view was that keyword campaigns at this scale were operationally impossible. Hundreds of thousands of properties, each needing its own keywords, maintained in real time as inventory changes. Early attempts confirmed it: Using property names directly from the feed produced keywords that rarely matched real search behaviour. Names were too long, inconsistent, or contained details that distorted intent. Ads couldn’t fit within RSA headline limits.
What we built
We built a system that evaluated each property individually using three inputs: the property feed (name, location, average booking value), Keyword Planner data mapped to each property, and AI filtering to remove queries that weren’t genuinely relevant – contact searches, restaurant terms, misleading amenity references embedded in property names.
Approved keywords were mapped to a structured intent dictionary. This let ATLAS generate consistent name variations for each property while keeping responsive search ads within headline character limits. Campaigns were structured around property value tiers, so bidding could optimise toward higher-value bookings rather than treating all inventory equally.
What it produced
- 350,000+ properties marketed
- 700,000+ keywords generated
- 39% of bookings from properties that had never converted before
- 49% of remaining volume from search terms that had never appeared in the account before
- CTR up 28% versus feed-based ads
- ROAS exceeded client targets by 15%, with 95% of bookings driven by exact match
The 39% figure is the one worth pausing on. These weren’t incremental gains on existing volume – they were bookings from properties that had never converted before. The feed-based setup wasn’t reaching them. Keyword structure did.
What this means before September
Accounts relying on DSA will be auto-migrated to AI Max. That can happen without any action on the advertiser’s part.
The question is whether advertisers want to take the risk that AI Max will perform identically straight out of the box. For accounts where DSA drives a significant proportion of total traffic, that is a substantial gamble. AI Max is a black box and there is no guarantee it replicates the same volume, efficiency, or coverage that DSA currently delivers.
Building a keyword-based alternative now provides a tested fallback – keyword structures that run alongside AI Max giving advertisers the control and visibility that automation alone cannot.
September 2026 is a fixed date.
Explore more case studies, or book a strategy session to talk through your account.





