I’ve spent most of my career in B2B, and at Made by ON I spend my weeks talking to leaders of the world’s most exciting B2B businesses. Lately those conversations keep circling one question, which I put to our strategists and UX team: what happens when buyers draft their request for proposal (RFP) evaluation criteria with an LLM?
Most industries are small and relationships count, so procurement of new services and tools for a business arrives in two temperatures. Friendly processes, where you’re known and your site’s job is to smooth the path and shape the requirement. Cold ones, now where a machine decides whether you exist at all.
Your website writes their RFP
The critical question is the website helping pass the screening criteria or is it shaping the brief?
Some fundamentals in business do not change, in bygone days, shaping the RFP was the oldest trick in enterprise sales. Get to the buyer early, help them define the requirement, and watch the requirement describe you. Now procurement teams are using AI to scan your site for as much of the information they need to construct the RFP before they even speak to you.
Nobody is searching for you - the cold truth of procurement
Semrush turns search visibility into evidence, the same kind of signal procurement teams now expect before a sales conversation starts.
Imagine the scene, a large RFP lands in your company’s shared email address cold out of nowhere. Palpable excitement from your sales team, and even though you might not win this time, it allows for a new set of contacts for the CRM.
This is gold dust because it shows that your website is working; digital is delivering top-of-funnel (TOFU) leads for the business. For us at Made by ON, it’s the biggest moment of validation in the investment cycle for a new B2B site, and we tend to celebrate this more than go live.
The latest research that is emerging is providing weight to this, and that’s because we need to step into the shoes of the procurement team; they’re working their business problem through an LLM, and somewhere in that exchange the machine suggests vendors.
Forrester surveyed 18,000 B2B buyers in January1. 94% used AI during their most recent purchase, and AI answer engines now outrank vendor websites and sales reps as the primary research source. 6sense tracked 4,000 buyers on deals with a median size above $300,0002; the winning vendor was on the day-one shortlist 95% of the time.
What this means for your organisation
Inside most organisations the website belongs to marketing, while sales depends on it to close. In stakeholder interviews that we’ve collectively spent the equivalent of weeks conducting, we’ve seen the same tensions surface every time: sales feels the site isn’t sweating hard enough on what they face in the field; marketing is trying to propel brand led growth strategies within the confines of the system created by various stakeholders from design, customer services, research teams and so on.
The first fix is change management, walking both sides through the anatomy of a B2B site as a machine reads it: the homepage, the product pages, the live content, and the tier of information nobody owns.
Homepage
Bluntly, machines don’t enter just through the front door; they land sideways, on whichever page answers the question. Therefore, every page is a front door now, and every page needs to say who this is for and why it holds up.
Product (PDP) pages
The product page has to work for both human buyers and the machine-readable evaluation layer forming around them.
This is where the battleground between sales and marketing exists.
In the machine world, the PDP stops persuading and starts evidencing. Enterprise procurement doesn’t want a benefits carousel. It wants specs, integrations, implementation timelines, commercial models, real pricing ranges.
Princeton ran the first peer-reviewed study of AI search in 20243, testing nine content tactics across 10,000 queries. Statistics, cited sources and quotations won, worth up to 40% more visibility in AI answers. The PDP that reads like a data sheet gets quoted, and the one that reads as a billboard gets skipped.
From a marketing and brand perspective, a balance needs to be found between the transactional machine world and the human eyes that need to feel that the product can solve a real-world problem for them.
Live content feeds e.g. blog, case studies, news
Case studies turn customer outcomes into evidence that both buyers and machines can use during evaluation.
The biggest wins (and ultimately success stories) are where we’re shaping the content strategy as part of the overall communications strategy, this is by stress testing the purpose and potential efficacy of the content.
Content written to feed last decade’s search algorithm is dead weight. Content that answers a buying-committee question, or builds the evaluation framework for your category, becomes the citation layer the machine draws from.
A cornerstone to a B2B site is case studies, this is ‘social proof’ without being in social channels and was critical when we developed ways for Rho (a leading financial platform) to handle competitor objections. Ultimately whether a machine reads the content or a human, the same questions need to be answered in an interesting and succinct way.
It has to cover the problem, what was built or solved, what it changed, and ultimately what benefit the end client had. The crucial aspect is to be confident and talk in numbers, have quotes and references that matter.
Security, compliance and partner information becomes part of the evaluation layer long before procurement speaks to sales.
And the unglamorous tier finally matters. Security posture, certifications, compliance, accessibility statements and a ton of relevant industry-specific information that is often invisible to the marketing/brand team.
Across all the touchpoints I’ve just covered, the information architecture becomes critical as product complexity increases, when clients reach Series B (in venture terms) or add multiple product lines (in more traditional industries), this is where our UX researchers come into their own solving extremely gnarly problems. That said and honestly, the problem today was the problem yesterday; semantic markup and structured data were machine-readable twenty years before anyone used the phrase.
Closing thoughts
When someone in your business asks you about AEO and GEO or ways that AI might be engaging with your services and products on your site, make it relevant; think about your sales process and think about your sales funnel.
Ultimately, whether you’re a marketer, sales, or technology leader, your website needs to help sales at the top, middle and bottom of the funnel, especially if you’re skewed to enterprise sales.
There is no doubt that website design and building is a scientific endeavour, but in the age of machines, the art is making it human too.
