AI Max Keyword Prioritization: What Changed for Exact Match (and the Fix)

You didn't touch your Search campaigns. They changed anyway.

Here's the scenario a lot of Indian D2C teams walked into over the past couple of weeks. Nobody edited the Google Ads account. Budgets were the same, creatives were the same, the target ROAS hadn't moved. And yet, by the first days of September, exact match queries that used to convert like clockwork looked softer, and a few founders noticed CPA drifting up on their best branded terms.

The cause wasn't the market or a bad week. It was a setting Google flipped on your behalf.

Starting Sept 1, Google began automatically upgrading Search campaigns that used campaign-level broad match or standalone automatically created assets into AI Max. If your campaign fit that shape, it moved, and you didn't get a yes/no prompt. Most of the coverage focused on the big headline — "AI Max is now on by default." The part that's quietly costing D2C brands money is smaller and more technical: how AI Max decides which of your keywords wins a given search.

A quick example of what "silent" looks like

Picture a skincare brand running a Search campaign with a broad keyword for "vitamin c serum" and a hand-built exact match keyword for "[brand name vitamin c serum]." Before the upgrade, a shopper searching the exact brand-plus-product phrase reliably triggered the exact keyword — the one with the sharper bid and the product-specific headline. After the move, that same search can be picked up by AI Max's matching instead, which may serve a broader ad with different assets and a system-set bid. Nothing broke, no error appeared, and the campaign still spent its budget. The founder only finds out when the weekly export shows the brand term converting a little worse than it did in August.

What "keyword prioritization" actually decides

Keyword prioritization is the rulebook Google uses when more than one of your keywords could match the same search. Say someone types your exact hero product name. In a classic Search account, that query could match a broad keyword, a phrase keyword, and an exact keyword all at once. Google doesn't fire all three — it picks one to serve.

The old default most accounts relied on

For years, the practical rule was reassuring: when a search was identical or near-identical to an exact match keyword you'd built, that exact keyword got the priority. That mattered because your exact match keywords are usually the ones you've hand-tuned — the tighter bids, the specific ad copy, the landing page that maps to that intent. You built them precisely so your most valuable queries served the way you wanted.

Why D2C margins live and die here

For a D2C brand running on thin contribution margin, the difference between "my branded exact keyword served" and "a broad keyword served instead" is real money. The exact keyword might carry a considered bid and a product-specific ad; the broad path might pull a wider net of queries, different assets, and a bid the system decides on its own. Same search, different economics. When you're defending a ₹450 CPA on your best-selling SKU, you don't want that decision drifting.

What Google changed on Sept 1

Alongside the auto-upgrade, Google spelled out a detail in its AI Max help documentation that's easy to miss. In plain terms, the guidance says that when you disable the broad match campaign setting or move a campaign to AI Max, your broad keywords are no longer treated as exact match keywords for the purposes of prioritization.

The line, translated

Read slowly, that sentence is doing a lot of work. Under the old broad-match-plus setup, some of your broad keywords could effectively behave with exact-match-level priority on the right queries. After the move to AI Max, that inherited treatment goes away. AI Max applies its own matching and prioritization logic, and it doesn't quietly promote those broad keywords to exact-match standing anymore.

So the search still gets an ad. Your budget still gets spent. But the internal decision about which keyword, bid, and asset combination wins that search has shifted — and for high-intent queries you'd carefully mapped, that shift is not always in your favour.

This is a documentation clarification, not a flashy new feature. That's exactly why it's dangerous: there's no big in-account banner shouting about it, so the change lands silently in accounts that were running fine.

The fix: re-add your exact match keywords

Google's own remedy is refreshingly blunt. If you want to keep your previous prioritization behaviour, add copies of your exact match keywords into the new AI Max campaign. Explicit exact match keywords still get exact-match prioritization inside AI Max — so by putting them back in, you re-anchor your most important queries to the keyword, bid, and copy you actually chose.

The tradeoff nobody puts on the slide

There's a cost, and you should know it before you bulk-upload anything. Those new keyword copies start with a clean slate. The historical performance data attached to your original keywords — the conversions, the learning, the quality signals built up over months — doesn't transfer to the fresh copies. In the short term, that can make automated bidding a little jumpier while it re-learns, and it can fragment your reporting if you're comparing week over week.

So this isn't a "do it everywhere" move. It's a "do it where it pays for itself" move.

Decide it per campaign, not per account

The pattern is simple to hold in your head: protect the queries you already know convert, and let AI Max roam on the ones you're still discovering.

A 30-minute audit before you touch anything

Before you start cloning keywords, spend half an hour getting the lay of the land. Rushing the fix into every campaign is how you trade one problem for a messier one.

Step 1 — Confirm which campaigns moved

In the account, look for Search campaigns that were running campaign-level broad match or standalone automatically created assets. Those are the ones eligible for the Sept 1 upgrade. If a campaign didn't use either, it likely wasn't auto-moved.

Step 2 — Pull a before/after on your money terms

Segment by your top branded and hero-SKU search terms and compare the two weeks either side of the upgrade. You're looking for queries where impression share held but conversion quality or CPA slipped — that's the fingerprint of a prioritization shift, not a demand drop.

Step 3 — Rebuild only the exact keywords worth protecting

For the campaigns and queries that fail the check, add exact match copies of just those keywords. Don't clone your entire keyword list out of anxiety; you'll drown your own signal. If your account is large and you'd rather audit and re-anchor keywords programmatically instead of by hand, that's the kind of repetitive, high-stakes cleanup a managed performance marketing partner should be doing for you every time Google changes the rules mid-quarter.

Step 4 — Watch the re-learning window

After you add the copies, give bidding a few days and resist the urge to yank levers. The new keywords need a little runway to gather data before the numbers mean anything.

What this signals for the rest of 2026

None of this is an argument to fight AI Max. That train has left — new legacy broad-match and auto-created-asset structures were already blocked from being created back in early August, and Dynamic Search Ads are lined up for their own migration in February 2027. AI Max is the direction Google's Search product is heading, and for a lot of D2C prospecting it genuinely works.

The lesson is narrower and more useful: automation is happy to make decisions for you, and it won't always tell you which ones. Your job shifts from setting the account to auditing it — knowing which queries are too valuable to leave on autopilot, and re-anchoring those by hand. Keep your brand terms, your best sellers, and your bottom-funnel intent under explicit control. Let the machine handle the wide, exploratory edges where you have less to lose. That division of labour is what keeps a thin-margin D2C account profitable while the platform keeps reshuffling the defaults underneath it.