Can AI replace a search (SEA) consultant?

Adjusting CPCs, writing three ads per ad group, adding exact match keywords: the typical week of a 2015 search consultant is now almost entirely automated. What remains is deciding who sets the instructions these automations follow, and who checks their results.

The 2015 week is gone

Take the week of a search (SEA) consultant ten years ago, on an ordinary e-commerce account. On Monday, they raised or lowered max CPCs keyword by keyword. On Tuesday, they went through the search terms report and added the best queries as exact match keywords. On Wednesday, they wrote two or three text ads per ad group to compare them. The rest of the time went into bid adjustments by device, hour and location.

Almost every line of that schedule is now a setting. Google states that with a Smart Bidding strategy (Target CPA, Target ROAS, Maximize conversions, Maximize conversion value) you no longer need to adjust bids manually, and these strategies only accept a -100% device adjustment. Since June 30, 2022, expanded text ads can no longer be created or edited. The responsive search ad that replaced them takes up to 15 headlines and 4 descriptions, which Google Ads assembles into different combinations depending on the search. On Search campaigns, AI Max extends matching beyond the keywords you bought, through broad match and keywordless technology.

Performance Max goes further. According to Google, a single campaign accesses all Google Ads inventory (Search, YouTube, Display, Discover, Gmail, Maps), and the bid strategy distributes spend across these placements based on the conversion goal. The consultant who used to pick placements one by one no longer has a setting to do so.

If you measure a search consultant by the time spent in the interface, the answer to the title is yes, and has been for several years. We listed the tasks the platforms already handle, and the right tool for the others, in our article on automatable paid media tasks.

A ROAS of 6 on brand searches

Take a hypothetical case. A furniture retailer replaces its Shopping campaign with a Performance Max campaign on Target ROAS, using revenue as the conversion value. After a month, the new campaign shows a ROAS of 6, well above the old one. The search terms report, which Google opened to all Performance Max campaigns in 2025, shows that a large share of conversions follow searches containing the brand name. These people already knew the site, and some of them would have bought through an organic result or the brand campaign.

The campaign did what its settings called for: generate conversion value while meeting the Target ROAS. But the bid strategy counts a sale caused by the ad and a sale that would have happened without it in exactly the same way: both are conversions attributed to the campaign.

The levers are in the interface. Brand exclusions stop Performance Max Search and Shopping ads from serving on brand searches. Negative keywords, and the lists that group them, can now be applied at Performance Max campaign level. And Google gives priority to Search campaigns when the query is identical to one of their exact match keywords. These settings are quick to apply. They still require a decision Google Ads does not make: should you pay for this brand traffic, and at what price? To answer it, we would start by comparing back-office sales before and after the exclusion, then run a test with a control group if the amounts justify it.

Where a search consultant's time goes in 2026

Upstream of the automations, the consultant decides what Google Ads optimises for: which conversion actions are primary, what value they carry, which Target CPA or Target ROAS matches the advertiser's break-even point. A mistake at this level ends up in every bid, and the bid strategy applies it as consistently as a good setting. The margin case, which Google Ads ignores until it receives the cost of goods sold, is covered in our article AI or paid media agency.

They also prepare what the automations assemble: headlines and descriptions for responsive ads, images and videos for asset groups, the product feed, landing pages. A responsive ad only combines the assets it is given, and AI Max adapts text from existing ads and pages.

Downstream, they read results against sources other than Google Ads (GA4, CRM, back office) and design the tests. A geo test is prepared before launch: regions, duration, and above all the result that would change the budget split. No platform sets that decision threshold on the advertiser's behalf.

What about AI agents?

Google and Meta now let you connect an ad account to an AI assistant. Google offers a read-only MCP server for the Google Ads API, and Meta offers connectors to manage campaigns from an assistant. A language model can read a search terms report of several thousand lines, classify queries by intent, write the weekly report or propose changes. The longest part of a consultant's analysis work can be handed to it.

Two limits remain. An agent connected to Google Ads reads the conversions Google Ads attributes to itself: if it recommends increasing the budget of the Performance Max campaign in the example, it relies on the same ROAS of 6. And an approval given in a few seconds no longer controls much, a risk we detail in our article on keeping human control over advertising AI.

This is the position we took with Pilot, the tool we are building. It analyses Google Ads, Meta and TikTok accounts and makes proposals, and its product page states that no change is applied without explicit approval. We design it to save the consultant the time spent on analysis, while leaving the decision with them.

Hiring a search consultant today

If you are hiring a search consultant, or assessing the one who runs your account, questions about CPC management and ad group structure no longer separate candidates. We would rather set a case: the export of a Performance Max campaign with its ROAS, its search terms report and back-office sales for the same period.

A good candidate starts by checking what the conversion measures (which action, which value, which conversion window), looks for the share of brand searches, then proposes a way to measure what the campaign actually brought in before touching the budget. If they immediately suggest raising or lowering the Target ROAS, they know the interface, and you still need to check that they can read an account.