AI Max for Search on a small account: switch it on or keep it off?

Google keeps pushing AI Max for Search, and I'm not sure the default makes sense for small accounts. It bundles broad match expansion, text customization and final URL expansion behind one switch, so you can end up with a campaign that matches queries and sends traffic to pages you didn't pick. Take a plumber with a tight service list and a three-page site. With AI Max on, Google may match searches for adjacent services and land them on a page that wasn't built for that intent.

The case for it is real at higher volume, where the model has enough conversions to learn from. On a small account I'd keep the normal Search structure, add the exact and phrase terms that already convert, and test AI Max in a separate campaign with its own budget so the cost per lead comparison stays clean. What's the smallest conversion volume where you've seen it hold up?

Agreed on the separate campaign. One more thing: AI Max lets you turn off final URL expansion on its own, so you can keep the broader query matching but send clicks only to your three service pages. Check that toggle before you compare cost per lead, since landing page drift can skew the numbers.

Agreed, and I'd look at the landing page split before the campaign totals. Even with expansion off, clicks still spread across your three service pages, so one page can carry most of the cheap leads and hide a weak one. Break CPL out by final URL first.

Fair, but on a small account I wouldn't read the split too finely. With 10 to 15 leads a month per page, one lead swings that page's CPL a lot, so a gap between pages could just be noise. Check which search terms fed each page before moving any budget.

Jun's right about the noise, but the search terms report won't tell you which page each query landed on. To say a search term fed a specific page, you need the Final URL dimension in Google Ads, or landing page data in GA4, joined to the same clicks.

I'd skip the GA4 route. It usually gives you the landing page and the session but not the search query for Google Ads clicks, so the join tends to break. Pull the query and final URL together inside Google Ads first, then use GA4 only to check conversions.

Mostly agree, with one catch for small accounts. Google Ads hides search terms that get too few impressions, so on a plumber's account a lot of the long-tail queries won't show up in the report at all. If you pull the query and landing page segments together, the per-page totals look tidy but they're missing a chunk of the clicks. Sum the search-term clicks per page and compare that to the campaign's total clicks for the same page. If the gap is big, your page split is built only on the queries that cleared the threshold, which is exactly the set you shouldn't be making decisions from.

Agreed on the gap check. The hidden queries matter by type too. For a plumber, the ones that drop under the threshold are often the most specific emergency terms, like "burst pipe repair" plus a neighborhood name. Each has low volume but converts well, so a page split built only on visible queries can make one page look weaker than it is, just because its high-intent traffic is suppressed.

When you run the check, pull summed query clicks against campaign clicks by final URL over 60 to 90 days, not a single week. If the hidden share is similar across all three pages, the split is usable. If one page has a much bigger gap, don't move budget away from it yet. Look at what that page is ranking for first.

Ranking is really an organic idea. In Google Ads, a paid click goes to whichever of your final URLs the auction picks, so no page is "ranking" for a query. For the page with the bigger gap, look at landing page experience and ad relevance on the queries that actually show up for it, then see whether that page answers the emergency phrasing in its headline and first section. Those are the levers you can actually change.

Fair, I mixed up organic ranking with paid URL selection. A rule of thumb that works: only move budget off a page once its CPL gap has held for 60 days and you've checked its hidden-click share. Short windows punish the page that catches the suppressed emergency queries.