The short answer: Forrester's 2026 survey of 18,000 global business buyers found 94% used AI during their most recent purchase process — 55% compared vendors inside AI tools, 54% researched product information there, and 47% built their internal business case before contacting any vendor. Buyers named generative AI and conversational search as a meaningful information source at twice the rate of anything else, ahead of vendor websites and sales contact. But Gartner found 69% still validate what AI told them with a sales rep. So AI is displacing discovery, not human decision-making. The practical consequence: your organic traffic can fall while your pipeline holds, and content now has to survive being extracted and summarised by something you cannot see.
If your organic sessions are down and your pipeline is not, you are not imagining it, and the explanation is now measured rather than anecdotal.
The two numbers that matter
Forrester's 2026 Buyers' Journey Survey covered 18,000 business buyers globally. The headline: 94% used AI during their most recent purchase process. Underneath it:
- 55% compare vendors against each other inside AI tools
- 54% research product information there
- 47% build their internal business case before contacting any vendor
And the finding that reframes the rest: buyers named generative AI or conversational search as a meaningful source of purchase information at twice the rate of any other source — ahead of vendor websites, product experts, and direct contact with sales.
Then the counterweight. Gartner's May 2026 survey found 69% of B2B buyers turn to sales reps to validate AI-generated insights, while a majority still prefer a rep-free experience across parts of the journey.
Read together, those two say something quite specific. Buyers are using AI to do the work that used to happen on your website, and then using a human to check whether the AI was right.
What this does to your funnel
The shape changed rather than the volume.
Under the old model, a buyer read three of your pages, downloaded something, got scored, and eventually took a call. You watched all of it. Under the current one, a buyer asks an assistant to compare four vendors in your category, reads a synthesis that may or may not include you, drafts an internal business case from it, and then — maybe — visits your site.
The change is easiest to see side by side:
| Old buying journey | 2026 buying journey | |
|---|---|---|
| Where research happens | Your site, review sites, analyst reports | Inside AI tools first, then verified elsewhere |
| What you can see | Most of it, in analytics | The late verification visit only |
| Comparison against rivals | On a review site or in a spreadsheet | Synthesised in a chat window you never see |
| Cost of being fourth-best | Fewer clicks | Absent from the answer entirely |
| Role of a salesperson | Guides the evaluation | Validates what AI already told them |
Three consequences follow:
Your first-touch data is wrong. The visit you record is not the first touch, it is a late-stage verification visit. That buyer formed an opinion somewhere you have no instrumentation.
Traffic and pipeline decouple. Sessions can fall while qualified conversations hold steady, because the low-intent research visits moved into a chat window. If you cut budget on a traffic chart alone, you will cut the thing that got you into the synthesis.
Being absent is worse than being fourth. In a list of ten organic results, position four still gets read. In a synthesised answer naming three vendors, four is invisible. This is the same winner-take-most compression happening in local search, arriving in B2B.
The buying group got bigger too
The same Forrester work puts a typical buying decision at 13 internal stakeholders and 9 external influencers.
That number changes what you should publish more than the AI statistics do. Your material is now read second-hand by people you will never speak to — pasted into a Slack thread, summarised into a one-page internal memo, quoted in a procurement document by someone in finance who has never seen your website.
Content that only makes sense in context does not survive that journey. Content built from specific, self-contained claims does.
What to do about it
Most of this is not new work. It is the same work, judged by a stricter standard.
Be specific enough to be quotable. "Significantly faster onboarding" cannot be cited by anything. "Onboarding takes 11 days on average across 40 implementations in 2025" can. Numbers, dates, named conditions. Vague copy has always been weak; now it is also unciteable.
Write passages that stand alone. Assume every paragraph will be lifted out of the page. That means naming the subject in the sentence rather than relying on "this" and "as mentioned above". It is the same discipline as writing a good abstract.
Answer the comparison question directly. 55% of buyers are comparing vendors inside AI tools. If you have no honest page about who you are and are not right for, the comparison happens entirely from third-party sources — review sites, competitor pages, forum threads. Publish your own version. Including the cases where you are the wrong choice, which is what makes the rest credible.
Be consistent everywhere. An assistant synthesising across your site, your directory listings, your LinkedIn and third-party mentions rewards agreement and gets confused by contradiction. Different positioning in three places reads as three weak signals rather than one strong one.
Make the site machine-readable. Server-rendered content, connected structured data, sane depth. The architecture side of this is unglamorous and it is the part most B2B sites fail on, because their marketing pages are client-rendered and effectively invisible to crawlers that do not run JavaScript.
Fix your attribution before you conclude anything. If 47% build a business case before contacting you, self-reported attribution on your forms is now more accurate than your analytics. Add "how did you hear about us" as a free-text field and read the answers monthly. That is the cheapest fix on this list and it is the one that stops you drawing the wrong conclusion from a traffic chart.
What has not changed
Some of the commentary around these numbers argues that content marketing is finished. The Gartner result says otherwise: 69% of buyers take what AI told them to a human to check.
Which means the job is unchanged in substance. You still need a clear position, material that demonstrates you have solved this problem before, and proof a sceptical person can verify. AI systems are a distribution layer on top of that, not a replacement for it. A content engine built on real expertise is what feeds the synthesis. There is no shortcut where you skip having something to say and optimise for the summariser instead.
The businesses that struggle here are the ones whose entire web presence could be regenerated from their category name. If a model can describe your company accurately without reading your site, you have not given it anything to cite.
Common questions
How many B2B buyers use AI in the buying process? Forrester's 2026 Buyers' Journey Survey of 18,000 global business buyers found 94% used AI during their most recent purchase. Within that, 55% compared vendors in AI tools, 54% researched product information, and 47% built an internal business case before contacting a vendor.
Does AI replace sales reps in B2B buying? No. Gartner's May 2026 survey found 69% of B2B buyers turn to sales reps to validate AI-generated insights, even while a majority prefer a rep-free experience for parts of the journey. AI is displacing early discovery, not the human validation step before a decision.
Why is my organic traffic falling while pipeline holds? Because the low-intent research visits have moved into AI tools. The visits you still record arrive later in the buying process, after the buyer has already compared vendors elsewhere. Traffic and pipeline have decoupled, so judging marketing on sessions alone now produces the wrong decision.
How do I get my company cited by AI tools? Publish specific, verifiable claims with numbers and dates in self-contained passages, keep your positioning consistent across your site and third-party sources, answer comparison questions directly on your own site, and make sure your pages are server-rendered with connected structured data so crawlers can read them.
Should I cut content marketing spend? Not based on a traffic chart. Check pipeline and self-reported attribution first. With 47% of buyers building a business case before contact, the material that shapes their decision is doing its work in a place your analytics cannot see.
The honest summary
Buyers moved their research into AI tools and kept their humans for verification. Your funnel now starts somewhere you cannot instrument and ends where it always did. Publish specific things that survive extraction, keep your story identical everywhere, ask people how they found you, and stop reading session counts as a scoreboard.
If your funnel data and your pipeline are telling different stories, that gap is usually the whole problem — the marketing systems work starts by measuring what is actually happening rather than what the dashboard reports.
Written by Jagatjeet — Jagatjeet (jagatjeet.com) is a digital marketer and web architect in Kamloops, British Columbia, working with founders on marketing systems and web architecture. Published 17 August 2026. Last updated 17 August 2026.