Google Drives Your Search Account Now. Here's What You Still Steer

Keywords are becoming suggestions. Google confirmed in 2026 that Dynamic Search Ads will sunset and auto-upgrade to AI Max, with the migration now starting in February 2027 after advertiser pushback bought a delay. Campaigns running broad match with Smart Bidding are on the same conveyor belt. Meanwhile the clicks you buy keep getting pricier: WordStream's benchmarks put the average search CPC at $5.26 in 2025, up 12.9% year over year and the latest in five straight years of increases. Compound a few years of that trend and you are paying roughly 45% more per click than you did in 2021.

So the machine sets the bids, picks the queries, and increasingly writes the ads. You can spend 2026 mourning match types or you can inventory what you still own. Five control surfaces remain fully in your hands. Each one below opens with the lever, the cost of ignoring it, and one setting to change today.

Your data feed is the bidding strategy now

The lever: conversion values, enhanced conversions, and Customer Match lists. Ignore it and Smart Bidding optimizes toward whatever junk you report, which usually means cheap, low-margin conversions at scale. The setting to change today: replace flat conversion counting with margin-adjusted or LTV-tiered values, then switch eligible campaigns to value-based bidding.

This is the one surface where your advantage compounds. Google cannot see your margins, your refund rates, or which customers reorder. You can. BCG found that digitally mature brands using first-party data in advanced marketing activations achieved 1.5x to 2.9x higher revenue uplift than brands that did not. Feed the algorithm profit signals and it chases profit. Feed it form fills and it chases form fills.

Start with the basics: enhanced conversions for leads, offline conversion imports from your CRM, and Customer Match lists segmented by value tier. If your first-party pipeline is thin, fix that before touching a single bid setting; our primer on what first-party data actually is and the first-party data playbook cover the collection layer.

Creative assets are your only voice left in the auction

The lever: responsive search ad assets, pinning, and AI Max's text guidelines. Ignore it and Gemini-generated headlines will speak for your brand, sometimes inventing offers you never made. The setting to change today: open the asset report, review every automatically created asset from the last 30 days, and remove anything off-brand. Then decide, ad group by ad group, whether text customization stays on.

AI Max's text customization generates new headlines and descriptions from your landing pages and existing ads. For most ecommerce accounts, that is free testing volume. For regulated verticals (finance, healthcare, legal), it is a compliance incident waiting for a screenshot. Google's own help documentation confirms you can set text guidelines and disable customization selectively, so use that granularity. Pin mandatory disclaimers to fixed positions. Write asset variants yourself for the claims that matter. The machine fills gaps; it should never define the message.

Landing pages decide where the traffic actually goes

The lever: final URL expansion paired with URL exclusions and url-contains rules. Ignore it and AI Max will route paid clicks to your blog archive, your careers page, and that out-of-stock product you forgot to redirect. The setting to change today: add URL exclusions for /careers, /support, /blog, and any thin or legacy paths, then whitelist your converting templates with url-contains rules.

Final URL expansion is the quiet half of AI Max. It lets Google pick the landing page it believes matches the query, which works well when your site architecture is clean and disastrously when it is a decade of accumulated URLs. Treat this like a crawl budget problem: audit which pages received paid traffic last month, kill the leaks, and keep the exclusion list living. Landing page conversion rate is still the multiplier on everything upstream, and the benchmarks in our piece on Google Ads conversion rates show how wide the gap between median and top-decile performers runs.

Brand controls are the new match types

The lever: brand inclusions, brand exclusions, and locations of interest. Ignore it and your prospecting budget quietly buys your own brand queries while your ads bleed into competitor-brand searches you cannot defend profitably. The setting to change today: on every prospecting campaign, set branded searches to unbranded-only; on brand campaigns, lock inclusions to your own brand list.

This matters more now because Google has folded brand inclusion and exclusion controls under the AI Max framework for all new Search campaigns, making the AI layer the gateway to core targeting. Locations of interest, set at the ad group level, are the other precision tool that survived: they let you separate "people in Austin" from "people searching about Austin," a distinction match types never handled cleanly. If you built your mental model around match-type hierarchies, our keyword match types guide explains how their behavior has already shifted, and brand lists are the successor system. Learn them now, while the migration is still voluntary.

Guardrails keep the machine honest

The lever: explicit tROAS and tCPA targets, budget caps, negative keywords, and a standing search-terms audit. Ignore it and "maximize conversions" without a target will spend every dollar you give it chasing marginal volume. The setting to change today: set an explicit ROAS or CPA target on every Smart Bidding campaign, even if it just codifies current performance.

Negatives still work under AI Max, and the search terms report got more useful: Google now surfaces the headlines and URLs served alongside each query. That transparency is only worth something if someone reads it. Book 30 minutes weekly, add negatives for irrelevant themes, and feed what you learn back into audience strategy; the segmentation logic in our audience targeting cheat sheet pairs naturally with this loop. And before you raise any target, run the math on what incremental volume actually costs you at the margin. A ROAS calculator takes two minutes and prevents expensive optimism.

What the results actually look like

Google's launch materials claim advertisers activating AI Max typically see 14% more conversions or conversion value at similar CPA or ROAS. Independent numbers are messier. An ALM Corp analysis of 250+ real campaigns found roughly 13% more revenue arriving alongside a 16% higher CPA. Both can be true. The gap between them is the gap between managed and unmanaged adoption, and the five surfaces above are the management.

Steer with what Google cannot see

Here is the honest read. Every control Google removes is one it believes its models handle better than the median advertiser, and for the median advertiser, Google is probably right. Your job is to stop being the median advertiser. The durable edge in 2026 lives in the inputs the auction cannot generate for itself: your margin data, your customer lists, your landing page quality, your brand judgment. Those five surfaces are small compared to what you controlled in 2019. They are also, dollar for dollar, more powerful, because everyone else is leaving them on default.

Sources

• Google: Dynamic Search Ads are upgrading to AI Max
• WordStream: Google Ads Benchmarks 2025
• BCG: Responsible marketing with first-party data
• ALM Corp: AI Max performance study, 250+ campaigns