AI agents and B2B buying: what changes for marketing teams

TL;DR

• B2B buyers now research with AI as a matter of routine, and they check what it tells them. That moves your website from the first stop to the place buyers and their tools go to verify.
• Five things change for marketing: the shortlist forms before the visit, proof outweighs claims, part of the buying group is software, the website becomes a validation step and agents start to act, not just read.
• The response is not a new channel to buy. It is plain, consistent facts on your own pages, measured in new ways, with someone on the team who owns how machines read the site.

What the research says, as of October 2026

Three large buyer surveys published in 2026 point the same way, even though their numbers differ:

• Gartner. A survey of 645 B2B buyers, run in August and September 2025 and published in May 2026, found buyers used an average of seven information sources in a recent purchase, and 45% used generative AI, mainly to research vendors and products. 69% prefer to validate AI-generated insights with a sales rep, and 51% think generative AI is the more likely source of misleading information (Gartner).
• Forrester. Its State of Business Buying 2026, based on a survey of nearly 18,000 business buyers, reports that "nearly all business buyers (94%) report using AI during their buying process," and that AI search tools "can create mistrust by delivering incomplete or unreliable information," so buyers seek validation from peers, experts and analysts.
• TrustRadius. Its 2026 report, based on a January 2026 survey of 1,862 technology buyers and 444 vendors, found 63% of buyers used AI during their purchase journey and 94% of those fact-check its responses at least some of the time. 83% shortlisted three or fewer products, and transparent pricing was buyers' top wish for the fourth year running (TrustRadius via HG Insights).

On the traffic side, Cloudflare reported in September 2026 that daily requests from AI agents on its network grew by more than 1,700% over a year (Cloudflare).

Where the sources disagree

The share of buyers using AI ranges from 45% to 94%, mostly because the surveys ask different questions: generative AI in one recent purchase, AI anywhere in the journey, or any AI use. Forecasts differ even more. In October 2025, Gartner predicted that by 2028, "90% of B2B buying will be AI agent intermediated" (Network World). That is a prediction, not a measurement, and it sits awkwardly next to Gartner's own 2026 finding that most buyers still want a person to confirm what the AI said. Plan for heavy AI-assisted research now and for agents that act on buyers' behalf as a growing, uneven trend.

Change 1: the shortlist forms before the visit

When a buyer asks an assistant for "three vendors that do X for companies like ours," the answer is a shortlist. If you are not on it, there is no visit to convert. That is zero-click search applied to vendor selection.

What it means for marketing:

• Your pages are read by machines first. The assistant's summary of your company is built from your pages and what others say about you. Our article on how ChatGPT, Gemini and Perplexity describe your company shows how to check yours.
• Category language matters. If your site never says plainly what category you are in and who you serve, an assistant has to infer it, and may put you in the wrong list.
• Measurement shifts. Google's Search Console added generative AI performance reports that show impressions of your URLs in AI Overviews, AI Mode and Discover. Google says they were rolled out to all websites worldwide as of August 31, 2026. Add them to your monthly report next to organic traffic.

Change 2: proof outweighs claims

Buyers who fact-check AI answers will fact-check your pages too. Forrester describes buyers as "far less forgiving of claims without proof." Slogans do not survive that check; specifics do.

• Publish the facts buyers ask for: pricing or how pricing works, plan limits, integrations, regions served, lead times, certifications and what you do not do.
• Make claims checkable. Link a security claim to the security page and a "works with" claim to the integration docs.
• Stay consistent across sources. A buyer who compares an AI summary, your website and a sales call will notice when they disagree. Every page, every listing and every sales deck should give the same answer to the same question.

None of this is AI-specific. Google's guide for its AI features says the advice that matters most is creating content people "find unique, compelling, and useful," and it warns that mass-producing pages for every possible query "violates Google's scaled content abuse spam policy" (Google Search Central).

Change 3: part of the buying group is software

Forrester puts the typical buying decision at 13 internal stakeholders and nine external participants, and finds that buying groups double in size, from seven to 14 members, when the purchase includes generative AI features. Procurement acts as a decision-maker in 53% of cycles. Each of those people may send an assistant to research you.

• Write for each role. Security wants data handling and certifications; finance wants pricing structure and terms; procurement wants supplier details. Give each a page an assistant can find.
• Make key facts machine-readable. Plain HTML text, accurate structured data and a short llms.txt file help tools read the facts correctly. Our guide to writing an llms.txt covers the file.
• Remove blockers. Login walls on documentation, PDFs with no HTML version and bot challenges on key pages hide exactly the facts procurement needs.

Change 4: the website becomes the validation step

If 69% of buyers want a person to confirm AI-generated insights, the visit to your site is often that confirmation step. The buyer arrives with an answer and a doubt.

• Answer the doubt fast. The pages buyers land on from AI answers, usually product, pricing and comparison pages, should state the key facts above the fold.
• Offer a conversation. An assistant on the site that answers from your own pages and cites them, with a quick path to a person, matches what buyers say they want: self-service first, a human for the moments that matter.
• Align sales and marketing. When a buyer says "ChatGPT told me you don't support SSO," sales should know where that came from and marketing should fix the page.

Change 5: agents start to act

Today most AI in B2B buying reads. The next step is agents that fill in forms, request quotes and book calls for a buyer. Agentic browsers can already operate pages, and the draft WebMCP standard lets a site offer agents named actions instead of making them guess at buttons; see WebMCP tools for B2B sites.

• Check that your forms work for an agentic browser: real labels, clear field names and no traps such as hidden required fields.
• Expect agent-submitted leads and decide how sales should treat them. A demo request filed by an agent for a real buyer is still a real lead.
• Watch your analytics for agent visits and referrals from AI assistants. They are a small share of traffic for most B2B sites today, so track the trend rather than the absolute number.

For the commerce side, including the protocols in development, read agentic commerce explained.

What to change in the marketing plan

A 90-day plan

What not to do

• Do not chase every new protocol. Most are drafts. Prepare what helps people and machines alike: clear pages and working forms.
• Do not block AI crawlers by reflex. Blocking assistant and search crawlers can take you out of the answers buyers read. Decide about training crawlers separately.
• Do not flood the site with AI-written pages for every query variation. Google treats that as spam, and buyers who fact-check will notice thin content.
• Do not hide pricing logic if your competitors show theirs. Transparent pricing has topped buyers' wish list for four years running in TrustRadius's surveys.

The common thread is generative engine optimization done honestly: GEO that works by making the facts easy to find, not by gaming the systems that read them.

AI agents do not replace B2B marketing. They raise the bar for the parts that were always important: being clear about what you do, being specific about what it costs and being consistent everywhere. Teams that treat their website as the source of truth for people and machines will show up on more shortlists and lose fewer deals to a wrong answer.