Search query reports used to be the foundation of every paid search account I built. Match a keyword, write an ad, point it at a page that repeats the keyword enough times to look relevant.
Google's test of conversational ad formats inside AI Mode and Gemini-generated responses breaks that whole sequence. Ads are starting to surface as part of an AI-generated answer, matched to the interpreted meaning of a conversation rather than a string of words.
The practical question that follows is uncomfortable for a lot of accounts: if Google is reading your webpage to decide whether your ad belongs in an answer, is your page actually saying anything worth advertising?
The mechanics here are worth sitting with. In a conversational format, Google's AI generates a response to a query, and your ad surfaces as part of that response. To do that, the system has to read your page and decide whether your content answers the conversation the user is having. Your landing page stops being a container for keyword density and becomes the brief Google reads before writing the ad.
That changes what a good page looks like. Someone searching "mortgage automation software" used to get matched on that exact phrase. Now the query that triggers your ad might be "What's the best way to automate mortgage document processing?" or "How do lenders reduce manual underwriting work?" The keyword is buried inside a problem. If your page front-loads the keyword and lists features, it has nothing to say to that conversation. If your page explains the workflow, names the pain point, and describes the outcome, the AI has something real to pull from.
So the work is about content depth, not keyword placement. The pages that will earn ad visibility do a few things well:
None of that is new advice for good content. What is new is that Google's AI is now the first reader, and it decides whether your ad shows at all.
If pages change, research has to change with them. The old motion was pulling a search terms report, sorting by volume, and aligning keywords to ad groups. That answers the wrong question for conversational matching. The new question is not "what do people type?" It is "what problem is this person working through, and how would they describe it out loud?"
That moves research much closer to building a marketing persona than running a keyword tool. I want to understand the pain points a buyer brings to a search, the way they phrase those pains, and the follow-up questions they ask once they get a first answer. Matching becomes about connecting a broader conversation and topic to a solution, not lining up exact-match phrases.
Keyword tools still have a place, but the richer inputs sit elsewhere:
Competitive positioning earns a new dimension too. It is worth watching how competitors get framed or recommended inside AI-generated responses, because that framing is a signal of which content themes Google trusts in your category.
All of this leaves a gap between a necessary shift in campaign management and the data Google currently reports. To optimize conversational ads well, I would want to see the prompts or question chains that triggered an impression, where in the AI response the ad appeared, and whether it surfaced as a recommendation, a citation, or a follow-up suggestion. I would want impression share inside AI Mode specifically, engagement tied to AI experiences versus traditional SERPs, and the follow-up query paths after someone interacts with an ad. Most of that visibility is not there yet.
The bigger measurement problem is attribution. A conversational session is multi-turn by nature. A buyer asks, reads, asks again, and your ad might assist a conversion three questions before the click that gets credit. Last-click and linear models will underreport what these interactions actually do. Strong attribution modeling stops being a nice-to-have and becomes the thing that keeps you from cutting campaigns that are quietly working.
Until Google closes that reporting gap, the practical move is to tighten what you can control: disciplined tracking, path analysis in GA4, and a habit of looking at assisted conversions rather than trusting last-click alone. Treat the back-end wishlist as a roadmap for the questions to keep asking your reps and your data, not as a reason to wait.
If you want a starting point, work in this order.
First, audit your top revenue pages against a single test: does this page demonstrate understanding of a buyer's problem in plain language, or does it just repeat a keyword and list features? Rewrite the ones that fail. Lead with the problem, the workflow, and the outcome.
Second, build a small intent map for your category. Pull 10-15 real conversational questions from sales calls, support tickets, and community threads, then group them into clusters by buyer stage. That map tells you what your pages need to answer.
Third, run a content gap check. Compare your intent clusters against what your pages actually cover, and fill the holes with FAQ-style content written the way buyers ask.
Fourth, start tracking which content themes and informational queries Google associates with your solution. That association is likely to shape ad matching inside AI Mode, so knowing it early is an advantage.
This matters most for complex B2B categories where buyers research in stages and phrase searches as problems rather than products. If your buyers run exploratory, problem-first searches, and your pages are currently optimized for keyword presence, you are exposed right now.
It is less urgent if you live in a highly transactional category where short, direct queries still dominate and intent is explicit. Someone searching for a specific part number with a credit card ready is not having a conversation. Even there, though, the behavior shift Google is responding to does not stop at your category line.
The advertisers who adapt fastest to Google AI Mode ads are the ones who are already treating landing pages as documents that prove they understand a buyer, and who have already done research that looks like persona work instead of volume reports. For them, this is less a pivot than a reward for established habits.
For everyone else, the cost of strong content and real intent research just went up, because Google is now reading the page before it decides whether you deserve to be in the answer at all. Bids still matter. They just stop being the lever that saves a page with nothing to say.
If you're running paid search and want to understand how your current landing pages and audience research hold up against Google's AI Mode ad matching, this is the right time to pressure-test your strategy with a team that's already working through these changes.
Google's AI reads your landing page content and decides whether it answers the conversational query a user is having, then surfaces your ad as part of the generated response. Matching is driven by how well your page demonstrates understanding of the buyer's problem in natural language, not by exact-match keywords. Content depth and topic coverage become the primary relevance signals.
Restructure pages around the problem you solve, the workflows involved, and the outcomes a buyer gets, instead of front-loading keywords and listing features. Add conversational, FAQ-style content that mirrors how people actually ask questions, and include clear expertise and trust signals. The goal is a page Google's AI can pull a genuine answer from.
Likely yes, if you rely on last-click or linear models. Conversational sessions are multi-turn, so an ad can assist a conversion several questions before the click that gets credit, which means standard models will underreport its impact. Lean on assisted conversion reporting and path analysis in GA4 now, and treat stronger attribution modeling as a priority rather than an afterthought.