
B2B buyers now build their vendor shortlist inside AI tools, before they ever reach your site.
94% of buyers used AI in their most recent purchase. Over half compared vendors and built business cases before contacting anyone.
AI Overviews cut clickthrough rate by 58% on top-ranking pages, up from 34.5% eight months earlier.
Non-brand search is where the cost lands. On average, B2B SaaS brands now pay $207 per non-brand lead against $34 for brand, with CPCs up 29% year over year.
Only 22% of marketers track AI visibility, which is why this is still an opportunity rather than a disadvantage.
This guide breaks down six shifts across discovery, search economics, audience building, and creative, with what to do about each and what to leave alone.
If you are not in the AI answer, you are out of the shortlist before your ads ever serve.
Transactional intent held, which makes this a reallocation problem, not a channel failure.
The cheap clicks are gone, so what remains is less forgiving of weak landing pages and loose match types.
Efficiency improves while volume shrinks, and most reporting only surfaces the first number.
Fewer top-of-funnel visits means smaller mid-funnel audiences, and it breaks quietly months before anyone traces the cause.
Buyers arrive educated, so explainer ads are wasted impressions.
Buying did not get more complicated. It got more private.
Research that once spanned eight browser tabs now happens inside one conversation. A buyer asks a question, gets a synthesised answer, asks a follow-up, and arrives at a shortlist. None of it touches your site. None of it appears in your analytics. By the time you see them, the comparison is done.
The funnel did not shrink. It moved upstream, into a place you cannot instrument and cannot bid on.
Forrester's 2026 Buyers' Journey Survey covered nearly 18,000 global business buyers. 94% used AI during their most recent purchase.
Not for admin. For the purchase itself:
All before contacting a vendor.
That last figure is the one to sit with. Almost half of buyers had written the argument for buying something before they spoke to anyone selling it.
A security director needs to replace an endpoint tool. Eighteen months ago that meant a Google search, four or five vendor sites, a G2 comparison page, two gated reports, and a demo request.
Now it starts with a question typed into ChatGPT. What are the best MDR providers for a mid-market company with a small security team?
Back comes a shortlist of four vendors, each with a line on who it suits.
Three prompts in, the shortlist is set. Two vendors have been described in a way that fits the requirement. Two have not. The director has not visited a single website.
When they finally do click through, it is to confirm a decision that has largely been made.
Marketing teams notice this late, because it surfaces first in places nobody flags.
The buyer who used to research by clicking now researches by asking. The clicks that formed the early journey stop happening. The informational query gets answered on the results page. The comparison happens in a chat window. What reaches your account is the tail end: people who already know what they want.
That produces a specific pattern. Fewer clicks, but better ones. Efficiency metrics improving while volume falls. Campaigns that look healthier in the platform than the pipeline feels in the CRM.
Search is not dying. Transactional intent is holding. But the part of the funnel where buyers used to click their way toward a decision has moved somewhere you cannot see.
Your ad account is where that first becomes measurable. It is rarely where anyone thinks to look.
The category question is increasingly being answered before a buyer ever reaches your site. AI summaries and conversational assistants are shaping how people understand a category, which vendors seem relevant, and what a reasonable shortlist looks like.
A buyer who needs to solve a problem may now ask a model to explain the category, name the options, and narrow them down. What comes back is often a compressed consideration set, and that can influence the rest of the journey before any direct interaction with your brand.
You are either in that early consideration set or you are not, and much of the evidence of that happens outside your ad account.
Bid strategy cannot reach someone who never entered the auction.
If a buyer's shortlist is set before they search, your non-brand campaigns are not always competing for first awareness. They are increasingly competing for confirmation among a smaller group who already think you are plausible. That changes what those campaigns are for and what you should expect them to deliver.
Pick 15 to 20 queries your buyers would actually type: category questions, comparison questions, and use-case questions. Run each through ChatGPT, Perplexity, and Google AI Overviews from a clean browser session, then log whether you appear, how you are described, and who appears alongside you.
Appearing in the list is table stakes; being described in a way that fits the buyer’s requirement is what helps you stay in the shortlist. Many brands will find at least one inaccuracy or a framing that quietly disqualifies them.
You cannot edit the answer, but you can improve the public material it draws on: your comparison pages, category pages, third-party listings, review profiles, and directory entries.
Generic positioning gives a model little to repeat, while concrete detail on who you serve, what you integrate with, and where you fit gives it something more useful to surface. Repeat the audit quarterly, because the answers change as models and competitors change.
Two people, a fixed prompt list, a spreadsheet, and a couple of hours a quarter is enough to start.
There is no reliable technical lever, and most of the tactics sold as one are just recycled SEO advice.
This shift tells you whether you have a visibility problem; the reallocation decisions come later, once you know where the money is actually going.
The questions that used to bring people to your site are increasingly being answered before they click.
“What is [category].” “How does [process] work.” “[Problem] causes.” These queries used to fill the top of B2B search programs. They were relatively cheap, they built awareness, and they gave you a first touch you could retarget.
That trade still exists, but it is weaker than it used to be. AI Overviews and other answer engines now resolve many informational queries directly on the results page, which means the impression still happens, but the click often does not.
When someone does click, they are more likely to be looking for detail the summary did not fully provide. That makes the traffic narrower, and often less predictable, than it was before.
This is the clearest reallocation opportunity in the guide, and one of the cheapest to act on.
You are not looking at a channel in decline. Transactional intent is still there. You are looking at a segment of your keyword list that may no longer justify the budget it receives, while the rest of the program continues to work.
This is a paid allocation problem, not a content problem. Those pages may still earn organic visibility and may be part of the source set AI systems draw from.
Ranking first on a query that gets answered above the fold often buys less than it used to.
Some informational queries still perform well, especially where the answer is complex, regulated, or not easily summarized. Segment first, then cut.
The cheap end of search is under more pressure, and non-brand is where that pressure shows up most clearly.
For years, non-brand search worked because the click was affordable relative to what it returned. A buyer researching the category clicked through, entered your funnel, and could be nurtured from there. Volume was high enough to absorb a lot of waste.
That buyer is now doing more research elsewhere, and what is left on non-brand terms is often a smaller, later-stage, more contested set of clicks. In that environment, the economics get worse even when the channel still works.
Non-brand is where a lot of discretionary spend sits, and it is the least forgiving segment when clicks get more expensive.
Every loose match type, weak landing page, or under-qualified keyword costs more than it used to because the click underneath it costs more. Precision that felt optional when volume was generous becomes much more important when the traffic is more expensive and more selective.
Blending them is how the problem stays hidden.
Broad match on non-brand terms may still have a role. However, it deserves much closer review when irrelevant clicks are materially more expensive.
A weak page wasted a cheap click before; now it wastes an expensive one.
Fund the top of that list properly and challenge everything ranked beneath it.
If leadership is still benchmarking non-brand against last year’s CPL without separating the channel mix, the discussion needs to happen before the quarter closes.
Later-stage buyers still search, and this is still where you can reach many of them.
Brand will usually look better on CPL, but that does not mean it is the best place to deploy all incremental spend.
Some of it is market pressure, not account quality, and loosening targeting to chase old click prices usually makes performance worse.
Branded search is often the best-looking line in most accounts, and that can mask a quieter problem: branded demand may be flattening or shrinking. Brand terms usually convert well and cost little because the searcher already knows the brand, but that has always been true.
What is changing is the volume behind those efficiencies. As buyers do more comparison and research inside AI tools and answer engines, some people who once would have searched your name to check you out may never do so. The ones who still search are often further along and more likely to convert.
That means conversion rate and cost per lead can hold steady or improve while branded impressions and clicks quietly fall. On the surface, the campaign looks healthy; underneath, the pool may be getting smaller.
Branded search is demand you captured, not demand you created. If branded volume falls, it usually points to weaker awareness, weaker consideration, or less upstream interest that later turns into a brand search.
Read as an efficiency win, a shrinking brand pool can lull teams into doing nothing. Read correctly, it is a signal about the stages before branded search, not a reason to celebrate the brand campaign itself.
Impressions and clicks over time are a useful demand signal, especially when viewed quarter by quarter.
If brand searches are falling while the category is flat or growing, that suggests share loss; if the whole category is losing momentum, that points to a broader upstream shift.
Navigational, research, and comparison-style brand searches can behave differently, and the mix tells you whether people are arriving decided or still evaluating.
If branded volume drops after awareness or category spend is cut, that is the relationship worth examining before you assume the brand campaign is the problem.
Bidding harder on your own name can move the metric, but it does not create new demand.
Falling branded CPL is not a win if it sits on top of falling volume.
This is a trend to monitor across quarters, not a week-to-week emergency. The right response is measurement first, then a read of what is happening upstream.
Retargeting depends on a steady supply of site visitors, and that supply can shrink when more research happens before a click. As buyers use AI tools and answer engines to get questions answered earlier, fewer people may reach the site, which reduces the size of retargeting pools downstream.
The core logic is sound: fewer clicks can mean fewer visits, fewer visits can mean smaller remarketing audiences, and the buyer who once landed on your site during research may now complete more of that research elsewhere.
The important correction is to frame this as a trend and a risk, not a universal certainty. Retargeting pools are not automatically shrinking in every account, but they can thin when top-of-funnel site traffic declines.
Retargeting is usually one of the more efficient parts of the account, so quiet erosion here can be expensive without being obvious.
A shrinking pool often does not announce itself loudly. The campaign may still return conversions, just fewer of them and at a gradually higher cost, which makes the decline easy to miss.
Pool size over time is the leading indicator; conversions are the lagging one.
Such as LinkedIn engagement, video views, event attendance, or community participation. Those sources can help reduce dependence on site traffic alone.
CRM lists, customer lists, and enrichment-based audiences are less exposed to declining site clicks.
Watch frequency closely so you do not simply over-serve the same people.
Rising frequency on a stable budget is often one of the first signs that the pool is contracting.
More money against a smaller pool usually buys more repetition, not more reach.
It still works; it is just more dependent on upstream traffic than it used to be.
Add one or two off-site sources first, then test whether they meaningfully offset the decline in site-based pool growth.
When the buyer arrives already educated, the role of the ad changes. A lot of B2B creative used to do the explaining work: what the category is, what the product does, and why the problem matters. That made sense when the ad was often the buyer’s first exposure to the idea.
Now that explanation often happens earlier, inside AI tools, answer engines, or other research surfaces. By the time someone sees your ad, they may already have a shortlist, so creative that simply re-explains the category is less useful than creative that proves why you are the better choice.
Many budgets still over-focus on audience settings, bid strategy, and placement, even as platform automation reduces how much control those levers provide.
The variable you control most directly, and the one increasingly shaping delivery, is the creative itself.
Outcomes, specific numbers, named results, and before-and-after framing usually do more work than broad category education.
If the buyer’s shortlist already contains two or three names, your ad needs to answer why you rather than just describe what you do.
On automated platforms, test more hooks, more claims, and more formats so the system can learn which messages pull the right people.
Teams adapting well are producing more variants and letting performance sort them instead of polishing a single concept for too long.
They have often already been given a framing of you, and creative that addresses that framing directly can outperform generic explainer copy.
On automated platforms, heavy manual layering can fight delivery rather than improve it.
Sharper messaging does not mean off-brand messaging. The goal is proof and differentiation, not shock value.
Some buyers still arrive early, and some categories still need education. The shift is toward more proof and less repetition, not zero explanation.
Six shifts only matter if you can see them in your own numbers.
This is the check: twelve questions grouped by shift, each with what a healthy answer and a warning sign look like.
None of this needs new software. Your ad platforms, your analytics, and an hour with a spreadsheet cover it all. Work through it once, and you will know which shifts are live in your account and which are not yet a concern.
Run 15 to 20 buyer queries through ChatGPT, Perplexity, and Google AI Overviews.
Healthy: you appear consistently and the description fits who you serve.
Warning: you are absent, or described in a way that quietly disqualifies you.
This is a qualitative check worth asking about.
Healthy: buyers arrive open.
Warning: buyers arrive with a shortlist you had no hand in.
Segment non-brand keywords by intent.
Healthy: the majority of spend is on commercial and transactional intent.
Warning: a large share still funds informational terms added years ago.
Compare quarter by quarter.
Healthy: stable.
Warning: falling while impressions hold, which suggests the answer is being absorbed before the click.
Never blended.
Healthy: you know both numbers and the gap is stable.
Warning: you only track a blended figure, which hides the split entirely.
Healthy: the two move together.
Warning: CPC climbs while conversion rate stays flat, so cost per lead drifts up.
Healthy: reviewed recently, negatives added regularly.
Warning: broad match is still running as if clicks were cheap.
Look at impressions and clicks over four quarters, not one.
Healthy: volume stable or growing.
Warning: efficiency is flat or improving while volume quietly falls.
Compare against category search trend.
Healthy: you are holding share.
Warning: your brand is falling while the category is not, which points to a share problem rather than a market-wide one.
Track pool size, not just campaign performance.
Healthy: stable or growing.
Warning: shrinking, with frequency rising to compensate.
Healthy: a meaningful share from CRM, engagement, and list-based sources.
Warning: almost entirely dependent on site traffic that is now thinning.
Audit your live ads.
Healthy: they lead with outcomes, proof, and differentiation.
Warning: they still explain what the category is to a buyer who already knows.
Count your warning signs.
You are in good shape.
Monitor the trends and repeat this quarterly.
The shift is underway in your account.
Prioritise the reallocation moves in Shifts 2 and 3, since those free budget immediately.
The pattern is well established and is likely already affecting pipeline.
Start with the diagnostic items in the search economics section. That's where the money moves fastest, then set time aside to work through the full set.
You do not need a transformation programme. You need to look at a few things in order, act on what you find, and leave the rest alone. Here is a sensible way to spend the next quarter, built entirely from moves in this guide.
Each phase is intentionally light. If a step shows that nothing is wrong in your account, skip ahead. The point is to find the shifts that are live for you, not to run all of them.
Take 15 to 20 queries through ChatGPT, Perplexity, and Google AI Overviews. Log where you appear, how you are described, and who appears beside you. Half a day of work can tell you whether you have a visibility problem.
Separate brand from non-brand everywhere, and pull 12 months of both. This single change surfaces most of what the search-economics shifts describe.
Informational, commercial, and transactional. You are looking for how much spend sits on informational terms and how those terms have trended.
Branded volume over four quarters. Retargeting pool size year over year. These are the two numbers that reveal the relevant shifts, and neither shows up clearly in a single month.
Move the freed budget to commercial and comparison intent.
Review match types, negatives, and landing page fit on your highest-spend terms. This is where an expensive click stops being wasted.
Take the two or three inaccuracies the audit surfaced and correct the public content the models draw on. Not the model, the source.
CRM, engagement, or list-based. Enough to see whether it holds up before you commit to more.
Shift the mix away from explanation and toward outcomes and differentiation. Brief a small batch of variants and let performance sort them.
If leadership benchmarks against last year’s cost per lead or a blended figure, have that conversation now. Use the split you built in month one.
Put the citation audit and the trend pulls on a quarterly cadence. The answers change as the models update and competitors improve their content, so a one-off audit ages quickly.
This guide is a read on a moving picture. Some of it is measured, and some of it is reasoned from what the measured parts imply, so it helps to be clear about where the edges are.
Several of the shifts in this guide are best treated as account-level hypotheses rather than universally settled rules. If your own numbers contradict them, trust your numbers.
If AI-referred traffic proves to convert as well as early signals suggest, the emphasis shifts from defending search toward earning citation, and citation share becomes much more important.
If platforms restore informational click volume, Shift 2 becomes less pronounced and the pruning advice becomes less urgent.
If retargeting pools hold steady despite falling site traffic, Shift 5 was overstated and should be revised.
Treat these as things to watch, not immediate action items. The shifts in the guide are the ones worth responding to now; these are the questions that will shape the next version.
The buyer has changed, and the account is where you notice it.
A lot of the research that used to happen on your site now happens elsewhere, earlier in the journey, and often through AI-assisted tools that compress the shortlist before a buyer ever reaches you. That can thin out clicks, shift intent downstream, and make the surviving metrics look better than the pipeline feels.
That is the trap: not obvious decline, but decline disguised as efficiency. A falling cost per lead on a shrinking pool, a blended number hiding a widening brand/non-brand gap, or a retargeting campaign starved by less upstream traffic can all look fine until you separate the numbers and read the trend lines.
You do not need a new stack, a new team, or a new budget. You need to look at a few things you are probably not looking at, act on the two or three that turn out to be real, and leave the rest alone.
Measure first. Reallocate second. Rebuild last, if at all.
The teams that come through this well are not the ones spending the most on AI. They are the ones separating brand from non-brand, pruning intent that stopped converting, checking how AI systems describe them, and moving existing budget toward buyers who still convert.
The window is open because most teams have not yet built a response. Measurement is still lagging buyer behavior, which means the gap between what is happening and what is being tracked is the opportunity.
Run the diagnostic. Find your live shifts. Reallocate against them. Then keep watching, because the picture is still moving, and the teams that keep measuring will keep adapting while everyone else is still arguing about whether any of this is real.
Need more help with your paid media campaigns?
Schedule a free, one‐on‐one session with me, and I’ll walk you through actionable steps to improve your campaigns and get better results.
2026 AI Buyer Effect: 6 Shifts Reshaping Paid Media