Something is changing about how buying decisions get made, and it is not gradual. AI agents acting on behalf of buyers are increasingly producing a shortlist - typically three to five options - rather than handing the buyer a set of search results to browse. The implication for Google Ads is uncomfortable but worth taking seriously: if you are not on that shortlist, you are invisible. Not ranked fifth or sixth. Not shown in position three with a lower click-through rate. Simply absent.
This is the impression squeeze. Total impressions available to a given advertiser in a given category may fall - not because competition dried up, but because the decision layer above the search results has thinned the field before any ad is served. Paid search teams need to think clearly about what this means for their accounts, rather than waiting to see what happens to their impression share reports.
Why Impression Share Tells You Less Than It Used To
Impression share has always been an imperfect metric. It tells you the proportion of eligible impressions you captured, but it says nothing about how the eligible pool was defined in the first place. In an agentic buying model, the eligible pool shrinks upstream of the auction. An AI agent researching a purchase on a buyer's behalf may surface a small set of providers based on structured data, availability signals, pricing information, and whatever the agent can verify - then proceed to transact or request contact without the buyer seeing a traditional search results page at all.
That means strong impression share within a reduced auction could be masking serious commercial risk. You might be winning 70% of impressions in your category while the total volume of those impressions has contracted sharply. Campaign-level metrics will not obviously signal this. You need to track absolute impression volumes over time, not just share, and watch for declines in click volume that are not explained by budget or bid changes.
What Gets You on the Shortlist
Agentic systems making purchasing decisions need something to evaluate. Where a human buyer might be influenced by ad copy, visual creative, or landing page feel, an agent is working with structured signals - pricing, availability, reviews, product specifications, and data it can pull and compare programmatically. That raises the stakes for how well your business data is exposed to automated systems, and how accurately it reflects what you actually offer.
For advertisers running Shopping or Performance Max campaigns, this reinforces the importance of feed quality in a way that goes beyond traditional best practice. Accurate pricing, complete product attributes, well-structured product titles, and up-to-date inventory signals are not just good hygiene - they become qualifying criteria for whether an agent considers your brand at all. An agent that cannot verify what you charge or whether you have stock has no basis to include you in its shortlist.
Conversion data matters here too. Smart Bidding and Performance Max already rely heavily on your conversion signal quality to decide where to compete. If your conversion tracking is incomplete, delayed, or polluted with low-quality leads, the bidding system forms an inaccurate picture of which auctions are worth entering. In an environment where agentic traffic may be higher intent but lower volume, wasting bid budget on the wrong signals becomes more costly, not less.
What AI Max and Performance Max Are Actually Optimising Towards
It is worth being precise about how Google's own AI-driven campaign formats interact with this shift. Performance Max and AI Max are not static systems - they learn from conversion data and adjust where they show ads based on what has worked historically. If agentic buyers represent a different behavioural pattern to traditional search users - shorter paths to conversion, less browsing, more decisive actions - the bidding system needs clean signal from those conversions to recognise and target similar opportunities.
This is a concrete reason to audit your conversion tracking now rather than later. Are you capturing micro-conversions that happen early in a compressed agentic journey? Are offline conversions being imported back into Google Ads with enough regularity to keep Smart Bidding current? If an agent-assisted buyer converts quickly and offline, and that conversion takes weeks to appear in your account, the bidding system may be undervaluing the auction type where those buyers first appeared.
AI Max's ability to generate search terms and match to intent beyond your exact keyword list also becomes more relevant in this context. If agentic queries look structurally different to traditional search queries - more specific, more transactional, potentially more conversational - relying solely on a tightly managed keyword list may exclude you from auctions you should be entering. That is not an argument for losing control of match types, but it is an argument for reviewing your AI Max and broad match settings with fresh eyes.
The Brand Signal Problem
Shortlist selection - whether by a human or an AI agent - favours brands that can be verified and trusted quickly. For paid search advertisers, this creates a tension that has always existed but now has higher stakes: you can buy clicks, but you cannot buy the underlying credibility signals that determine whether an agent includes you in its consideration set.
Review volume, review recency, and the consistency of your business information across platforms all feed into the verifiability of your brand. Advertisers who have treated these as secondary concerns - something for the SEO team or the customer service team - need to reconsider. If your Google Business Profile is stale, your reviews are sparse, or your pricing information is inconsistent across your website and your feed, you are creating gaps that automated systems will either fill with uncertainty or use to exclude you.
This is not about gaming a new ranking factor. It is about recognising that paid visibility and brand credibility are becoming more interdependent. An advertiser with strong brand signals and good conversion data will perform better in AI-mediated auctions than one relying purely on bid levels and creative. The bid is still necessary, but it is no longer sufficient on its own.
Practical Steps for Paid Search Teams Right Now
Start with your measurement foundation. If your conversion tracking has gaps - untracked form submissions, delayed offline import, missing micro-conversions - fix those before adjusting bids or campaign structure. Smart Bidding can only optimise towards what it can see. In a shortlist environment, the quality of your conversion signal matters more than the quantity of your keywords.
Review your product and service data for accuracy and completeness. For e-commerce advertisers, that means your Merchant Centre feed. For lead gen advertisers, it means your landing page content, your ad extensions, and the structured information you expose about pricing, services, and eligibility. If an agent cannot quickly extract reliable answers to the questions a buyer would ask, you are at a disadvantage before the auction even starts.
Finally, treat absolute impression volume as a metric worth monitoring, not just impression share. Set up a regular check in your GA4 or Looker Studio reporting to watch total paid search impressions and clicks by campaign type over time. If you start seeing compression that is not explained by budget or seasonal factors, you have an early indicator that the shortlist dynamic is affecting your category - and you will be better placed to respond if you spotted it early rather than after revenue had already softened.