Check your Google Ads account. There is a decent chance a campaign you built three years ago has quietly been converted into something else while you were doing actual work.
Google has been auto migrating older Search setups into AI Max. Microsoft went further and made its equivalent the default for new Search campaigns.
You can delay this. You cannot stop it. The platforms have decided that keyword level control belongs to a machine now, and the only real decision left to you is which levers you hold onto while the rest goes automatic.
Get that decision wrong and you will spend six months buying leads you cannot use.
What AI Max actually does to your account
Strip away the branding and it is three changes bundled together.
It broadens matching well past your keywords, using the intent of the query and the content of your landing pages rather than the terms you selected. It writes and assembles headlines and descriptions on the fly. And it decides which URL to send each query to, sometimes overriding the one you specified.
In other words, it takes over the three jobs that used to define search advertising. What you target, what you say, and where you send them.
For some accounts this is a genuine gift. If you sell one thing to a broad market and every enquiry is roughly equal in value, handing over the mechanical work usually improves results and frees up your week.
For others it is a slow disaster, and the pattern is predictable.
Automation optimises toward the target you set. If your target is "a form submission" it will find you an infinite supply of people who fill in forms and never buy.
The four levers that must stay yours
One. The conversion definition. This is the whole game and it is where most accounts lose control. If the system is optimising for raw lead volume, it will deliver raw lead volume, cheerfully, at a lower cost per lead than you have ever seen, and your sales team will spend three weeks talking to people who wanted a free quote for something you do not sell.
Feed it qualified events instead. A booked appointment. A quote accepted. A sale, with value attached. Offline conversion imports exist precisely for this and almost nobody uses them. Do that and the automation becomes an asset, because it is now hunting for the thing you actually want.
Two. Budget and the pace of change. Automated systems learn by spending. That is fine when you have decided how much tuition you are paying. Set the budget deliberately, change one thing at a time, and give each change enough runway to produce a readable result. Constant fiddling resets learning and guarantees you never find out what worked.
Three. Negative terms and brand safety. Broad matching pulls in queries you would never have chosen. Some of those are gold. Some are people looking for free advice, DIY instructions, jobs, or your competitor's warranty department. Review the search terms report weekly for the first two months and build the negative list aggressively. This is the single highest value hour in the account.
Four. The message and the landing experience. The system will generate combinations of your assets. It will not invent your positioning, your guarantee or your offer. Those are yours. Give it strong raw material and it assembles something good. Give it generic material and it assembles something generic, faster.
The migration checklist
If a migration is coming, or has already happened, work through this before you judge the results.
- Fix your conversion tracking first. Nothing else matters if the signal is wrong. Confirm what counts as a conversion, remove duplicates, and stop counting page views and phone clicks as if they were sales.
- Import your real outcomes. Push qualified leads and closed revenue back into the platform. This is the difference between automation working for you and working on you.
- Clean your landing pages. The system now reads them to decide what to match against. A vague page produces vague matching. Make each page unambiguous about what it sells and to whom.
- Set up a search terms review ritual. Weekly, in the diary, for eight weeks. After that, fortnightly.
- Keep one control. Where budget allows, hold a tightly controlled campaign on your highest value terms so you have something honest to compare against.
The branded search question
One specific trap worth naming. Broad automated matching loves branded searches, because they convert brilliantly and make the whole campaign look like a triumph.
They were also going to convert without you paying for them. If you are not separating branded from non branded performance, the automation will happily spend an increasing share of your budget buying traffic you already owned, and the dashboard will glow.
Split it out. Judge the two separately. This alone recovers real money in a lot of accounts.
The honest verdict
Automated search is not a scam and resisting it entirely is not a strategy. On clean signal, with a decent offer and strong landing pages, it genuinely outperforms most manual account management. The machine is better than you at the mechanical parts, and it never gets tired at 4pm on a Friday.
What it cannot do is decide what a good customer looks like, what you are willing to pay for one, or what makes you worth choosing.
Hand over the mechanics. Keep the judgement. That division of labour is the entire skill of running paid media from here on, and the accounts that get it right will quietly outperform the ones that either fought the change or surrendered completely.
Go and look at what your account is optimising toward today. If the answer is "form fills," you have found your afternoon.