Bot traffic overtook human traffic for the first time in June 2026, and a fast-growing share of it is AI agents researching purchases on buyers’ behalf. For MSPs and ISVs, that reorders the buying journey: an agent now sifts the field and draws up a shortlist before a human ever lands on your site. Your next customer’s first impression may be formed by a machine. What wins those agents over is structural, specific, evidence-rich content, and the playbook for it already exists.
Sometime in early June 2026, the internet crossed a line. According to Cloudflare, bots now generate more web traffic than humans — 57.5% of all requests, a milestone the company’s CEO didn’t expect until 2027. And the fastest-growing slice isn’t the familiar crowd of search crawlers and scrapers. It’s AI agents: tools like ChatGPT, Copilot and Claude, browsing the web on behalf of real buyers.
Which means there’s a reasonable chance the next “visitor” researching your MSP or evaluating your ISV product isn’t a person at all. It’s an AI agent, sent by a person, gathering evidence before they ever click through themselves.
If that sounds like a consumer-shopping problem, the research says otherwise. The B2B numbers are, if anything, starker.
What the research shows: B2B buyers now trust AI tools over your website
B2B buyers have made AI their primary research source — ahead of vendor websites and sales teams. Forrester’s Buyers’ Journey Survey, published in January 2026, found that 94% of business buyers now use AI in their buying process, and twice as many named generative AI or conversational search as a more meaningful source of information than any other, ahead of vendor websites, product experts and sales teams. A year after first reporting the trend, Forrester’s analysts observed that buyers “have since turned their use of AI up to eleven.”
Look at where that puts everyone. Buyers rate a generative AI tool as a more meaningful source than the website you built or the salespeople you pay. The shortlist is being drawn up out of sight, and an agent may have weighed you against three competitors before a human ever arrives.
Gartner’s research points the same way. Its survey of 646 B2B buyers, conducted in late 2025, found 67% prefer a rep-free experience, and 45% used AI during a recent purchase. “Sellers can’t rely on static collateral to carry influence in those moments,” warns Alyssa Cruz, Senior Principal Analyst in the Gartner Sales Practice. Gartner’s prescription is telling: structure content into modular, agent-ready building blocks that can be assembled into context-aware resources on demand.
There’s nuance worth holding onto, though. Forrester’s State of Business Buying, 2026 report found that while genAI searches are the starting point, buyers still lean on their internal and external networks — now averaging 13 internal stakeholders and nine external influencers per decision — to validate and de-risk what the AI tells them. That means the human relationship isn’t dead. But increasingly, the AI picks who buyers even talk to.
What AI agents want: inside OpenAI’s and Google’s new content specifications
We no longer have to guess what AI agents want from web content, because the platforms have written it down.
In late 2025, OpenAI and Stripe published the Agentic Commerce Protocol, including a Product Feed Specification that tells merchants exactly how to structure data for discovery inside ChatGPT. Google followed in January 2026 with its Universal Commerce Protocol, and announced dozens of new Merchant Center attributes for conversational commerce at I/O in May 2026.
These specifications were written for retail, but they amount to a manifesto on what machines find persuasive:
Real questions, really answered.
OpenAI’s spec includes a dedicated q_and_a field. Google’s new conversational attributes cover answers to common product questions — so an agent can respond to “is this jacket waterproof?” without guessing.
Evidence, not adjectives.
OpenAI asks for raw review data — actual customer sentiment, not marketing summaries — because user-generated insight is what lets an agent recommend with confidence.
Use-case context.
Both specs reward data that explains what something is for, who it suits, and under what constraints — the connective tissue between a buyer’s stated need and your offering.
What’s absent from both specifications is just as instructive: superlatives, slogans, “industry-leading” anything. The machines want structure, specifics and proof.
How Microsoft MSPs and ISVs should respond: five principles for agent-readable content
Microsoft Partners don’t publish product feeds per se (though ISVs with AppSource listings are closer to this world than they think). But the commerce principles translate directly to services and software businesses — and they expose a weakness on most Partner websites.
1. Look past your designations.
When an agent researches “which Microsoft Partner can migrate us to SharePoint Online before end of support?”, it hunts for information that differentiates. Solutions Partner designations, certification badges and “Microsoft 365, Azure, Security” service lists don’t do that job — every serious Partner has them. They prove you’re in the category. They give an agent nothing to choose you within it.
2. Lead with evidence, not claims.
Compare two statements an agent might find. “We are experts in Dynamics 365 migrations” — a claim every competitor also makes. “We moved a 300-user engineering firm from AX 2012 to Dynamics 365 Finance & Operations in nine months, with no disruption to live production scheduling.” It has evidence: specific, checkable, matchable to a buyer’s actual situation. This is precisely why the commerce specs demand review data over descriptions. Case studies with named outcomes are the B2B equivalent.
3. Apply the Q&A test.
The developers behind our own site Moretag, who’ve been preparing ecommerce clients for these specifications, use a quality test we’ll happily steal for B2B: could a buyer answer this question by reading your existing spec sheet — or, in Partner terms, your services page? If yes, the Q&A adds nothing. “Do you offer Microsoft 365 support?” fails the test. “Will migrating from SharePoint 2016 break our existing Power Automate flows?” passes it — because answering requires genuine expertise, and that’s exactly what agents (and buyers) are searching for.
4. Be specific about who you’re for.
Moretag calls use-case descriptions “intent recognition anchors” — the phrases connecting a buyer’s stated need to your offering. The test: would this phrase help an agent choose you over others in the category? “We serve businesses of all sizes across all industries” fits everyone and matches nothing. “Dynamics 365 for mid-sized UK manufacturers running ageing AX environments” matches a real query a real buyer will type. Agents don’t reward breadth. They reward fit.
5. Say where you don’t fit.
One of Moretag’s more counterintuitive findings: the products performing best in LLM-based discovery are honest about the situations where they fall short. For a Microsoft MSP Partner, “we’re not the right fit for sub-20-seat businesses” reads as a trust signal — to the agent, and to the buyer scanning its summary.
The practical checklist for Microsoft Partner marketing teams
Audit your site as an agent would.
Strip away the design and read the words. Could an AI distinguish you from your three nearest competitors? If every claim could be swapped onto their sites without anyone noticing, no amount of traffic will help.
Turn case studies into evidence.
Sector, scale, timeline, outcome. “Delighted with the service” matches nothing.
Build Q&A content from real buyer questions.
Interview your consultants, your support desk, your sales team. Publish what prospects actually ask, with honest answers. Apply the test ruthlessly.
Tighten your positioning.
Name your sectors, your typical client size, your designations, specialisms — and who you’re not for.
Mind the technical layer.
Schema markup, descriptive ALT text, clean marketplace listings. For ISVs, your AppSource listing is now agent-readable shop frontage; treat it accordingly. And remember Gartner’s advice: modular content blocks that agents can assemble beat monolithic brochures.
The web has changed. The discipline hasn’t.
The content that performs best with machines is the content that was always best for humans — that’s what strikes us most, reading the research and the specifications side by side. Grounded in real expertise. Specific rather than generic. Honest about limitations. Built from evidence, not assertion.
The bots took over the internet last week. For Microsoft Partners who’ve been doing content properly, nothing about the job description changed — the audience just got bigger, and considerably more literal.
What would an AI agent say about your business today? It might be worth asking one.
And if you don’t like the answer? The fix is usually in the words. Bright Star works with Microsoft Partners to build the evidence agents are looking for case studies with real outcomes, Q&A drawn from your consultants’ actual expertise, positioning sharp enough to match a real buyer’s query. Journalist-written, interview-led, and Co-op claimable. Let’s talk.
(And yes — we restructured this article using the same principles it describes. Descriptive subheadings, self-contained sections, and the Q&A below. Well, it would have been rude not to.)
Frequently asked questions
Do AI agents actually influence which Microsoft Partner a buyer chooses?
Increasingly, yes — before any human contact. Forrester found 94% of B2B buyers use AI in their buying process, and rank it as a more meaningful information source than vendor websites or sales teams. Buyers still validate AI findings with colleagues and external experts, but if an agent can’t find evidence that differentiates your business, you may never reach that validation stage.
Does this replace SEO for MSPs and ISVs?
No — it extends it. Schema markup, clean site structure and descriptive ALT text serve both traditional search and AI agents. What changes is the content standard: agents reward specific evidence, real Q&A and clear positioning over keyword-optimised generalities. Work that improves your AI visibility will almost always improve your human credibility too.
What’s the fastest first step for a Microsoft Partner marketing team?
Run the agent test yourself: ask ChatGPT, Copilot or Claude to compare your business against two named competitors for a service you offer. The gaps in what the AI can say about you are your content priorities — usually missing evidence (case study specifics), missing answers (real buyer questions) or missing positioning (your ICP, i.e., who exactly you serve).
Thanks to our friends at Moretag for the inspiration for this piece.
Read other articles by Juliet Stott and connect with her on LinkedIn
