AEO
Running Paid Ads When AI Is Doing the Pre-Search
November 18, 2026 · 5 min read · AEO
Your paid costs are up and conversion is flat. The cause is usually account-level, not market-wide. Here is how to rewrite ads for buyers AI already shortlisted.

Your paid search costs are climbing and conversion is not moving, and the reason is probably not that clicks got more expensive. The reason is that the buyer decided before the auction ran, and your ad is still written to introduce you.
That is the shift. The ad's job used to be discovery. Now it is confirmation. More of your clicks arrive from people who already asked ChatGPT or Perplexity or Gemini who the top options are, already got an answer, and are clicking to verify it. If your headline is still explaining what your category is, you are paying a premium price to re-litigate a decision that was made somewhere you were not watching.
Why your ad costs are up when the benchmarks say CPCs are flat
Start by separating price from yield, because most teams blame the market and then reallocate budget for the wrong reason.
The market story is rarely what the panic suggests. Before you conclude the auction got more expensive, check whether your own price per click actually moved, and whether it moved on brand terms, non-brand terms, or both. Those are three different problems with three different fixes.
If your costs are up, the cause is usually inside your account. Three usual suspects:
- Brand defense creep. You expanded brand coverage or let broad match bleed into it.
- Match type expansion. Non-brand terms absorbing budget that converts at a fraction of brand.
- Shrinking organic click share. The free click that used to sit above your ad is gone, so you are buying traffic you used to get for nothing.
Diagnose first. Pull brand and non-brand CPC and conversion rate as separate lines, not as a blended account average. A blended number hides exactly the thing you need to see.
The shortlist is built before the click
Here is the behavioral change that reframes everything else. When an AI answer sits above the results, fewer people click through to a traditional listing, and the links inside the answer do not absorb the difference. On a growing share of searches, the paid slot is the last clickable thing on the page.
Meanwhile, the volume of direct referral traffic from AI tools stays small. Most teams read that small number and dismiss the whole thing. That is the misread. The influence does not arrive as AI referral traffic. It arrives as branded search, direct visits, and higher-converting paid clicks from people who formed an opinion off-site. The shortlist is invisible in your reports because it happened in a channel that does not send you a session.
The practical consequence: generic queries now carry browsing. Branded and comparison queries carry the decision. Price and copy accordingly.
Your ad is answering a question the buyer already answered
Confirmation-stage copy does one of two jobs: it validates what the AI answer said about you, or it corrects it.
That is a different writing assignment than category education. Stop spending headline characters on what the product is. Spend them on what a finalist is checking.
- Lead with the specific claim, not the category. Named capability, named outcome, named number. The buyer already knows what you sell.
- Make pricing legible. Visible price, a price range, or the logic ("starts at," "based on ad spend," "fixed monthly"). Finalists are filtering on economics, and ambiguity reads as risk.
- Pre-empt the comparison. If the AI answer frames you against two alternatives, the ad and the page should address the dimension that comparison turns on, not avoid it.
- Correct what is wrong. If AI answers describe you with outdated positioning, an old service line, or a competitor's framing, the ad is your fastest correction surface while the slower citation work compounds.
- Cut the education. You are paying a premium CPC to teach someone who was already taught. Move that content to where it earns citations instead.
What changes on the landing page
Decision-strength pages, not discovery pages. Confirmation-stage visitors arrive to verify, not to learn, so the page has one job: close the open question fast.
Visible pricing or pricing logic above the fold. A comparison block that names the real decision criteria. Proof that is specific and checkable, not adjectives. Schema, structured comparisons, and clean FAQ architecture so the same page is also readable by the engines forming the next shortlist. One obvious next step, not five.
That last one decides the rest of the page. A page built for a browsing visitor offers options. A page built for a finalist offers a decision.
Is branded search a cost or a scoreboard?
Branded search is your demand-creation scoreboard. Teams keep treating brand clicks as credit-stealing waste and try to cut them. Wrong instinct.
When AI answers, podcasts, earned media, and video do their job, the result shows up as people typing your name. That rising branded volume is the measurement of work that has no clean attribution path. Defend the term deliberately, watch the volume trend as a demand metric, and judge non-brand spend on whether it feeds that trend.
What to measure each month
Paid performance and AI visibility are usually reviewed separately, by different people, in different decks. That is why neither one explains the other. Read them as one system, monthly:
- Brand and non-brand CPC and conversion rate, split, never blended.
- Branded query volume trend, treated as a demand-creation output.
- AI referral sessions and their conversion rate, tracked even at tiny volume.
- Citation presence across ChatGPT, Perplexity, Gemini, and Claude, and what those answers actually say about you.
- The budget rule that follows: protect confirmation-stage capture, pull spend from educating buyers who were already educated.
The link between those reports is the part most teams cannot see yet. The overlap between what ranks first on Google and what AI engines actually cite has collapsed from roughly 70% to as low as 17% in 2026. Ranking first does not make you the cited answer, and being the cited answer is now what sets your paid conversion rate before the click happens.
That is the whole argument. Your ads are not underperforming because the auction got expensive. They are underperforming because the decision moved upstream and your copy did not follow it.
We run performance marketing and AI visibility optimization as one discipline, because the shortlist and the click are one system. If you want to know what AI engines are currently telling your buyers about you, that is where an AI visibility audit starts.
Frequently asked questions
Are AI Overviews the reason my CPCs went up?
Usually not at the market level. Tinuiti's Q1 2026 benchmark data showed brand keyword CPCs down 9% year over year after four quarters of deceleration, with overall average CPC growth flat and clicks up 14%. If your costs are rising, look at account-level causes: brand defense creep, match type expansion, or shrinking organic click share on queries where an AI answer now absorbs the free click.
AI referral traffic is a rounding error for us. Why should I care?
Because the influence does not arrive as AI referral traffic. WebFX found generative AI referrals at 0.18% of sessions in 2025 while converting around 1.2x organic search. The shortlist forms off-site and then shows up as branded search, direct visits, and paid clicks from people who already have an opinion, so the effect gets credited to the wrong channel.
What does confirmation-stage ad copy actually look like?
It validates or corrects what an AI answer already claimed about you. Lead with a specific capability, outcome, or number instead of category education. Make pricing legible, address the comparison dimension the decision turns on, and cut anything that explains what your category is.
Should I stop bidding on my own brand terms?
No. Branded search volume is your demand-creation scoreboard, the visible output of work with no clean attribution path. Defend the term deliberately, track the volume trend as a demand metric, and judge non-brand spend on whether it feeds that trend.
How do I report paid and AI visibility together?
One monthly read with five lines: brand and non-brand CPC and conversion split, branded query volume trend, AI referral sessions and their conversion rate, citation presence across ChatGPT, Perplexity, Gemini, and Claude, and the budget rule that follows. Reviewing them in separate decks is why neither report explains the other.
