BOFU: Does AI Recommend You, or the Competitor?

This is the moment the buyer asks AI to choose. Not "what are my options" or "how do they compare," but "which one should I go with?" The answer often names a single brand. Sometimes it names two with a clear preference. Either way, this is where AI visibility turns into revenue, or into a competitor's revenue.

What AI is doing at this stage

AI is making a recommendation against the buyer's specific constraints. By now the buyer has told it a lot: the category, the region, who they are, what they need it for, and often their budget, timeline, and which competitor they're weighing you against.

AI has to check each of those constraints against each candidate. Does this provider serve that region? Does it handle that use case? Is it priced in range? Can it start by that date? Every constraint AI can confirm for a brand makes that brand safer to recommend. Every one it can't confirm becomes a hedge, or a reason to pick someone else.

That is the core dynamic of BOFU: AI recommends the brand it can verify, not necessarily the brand that fits best.

The prompts buyers actually type

These are the five BOFU templates SurfaceMap runs. Four name your brand. One deliberately does not.

ID	Shape	Question	What it tests
B1	Named recommendation	Would you recommend (Brand Name) to (Target Audience) choosing among (Category) options for (use case)?	Whether AI will vouch for you directly, and with what caveats
B2	Unbranded best-choice	Who is the best choice for (use case) among (Category) providers in (Region)?	Whether you get picked when the buyer doesn't name you. This is the most important question in the set.
B3	Value / pricing	Is (Brand Name) a good value choice for (use case) compared to other (Category) options in (Region)?	Whether AI can find and defend your pricing and value
B4	Urgency / readiness	If (Target Audience) in (Region) is ready to hire a (Category) provider for (use case) right now, would (Brand Name) be a suitable choice?	Whether AI can confirm availability, fit, and ability to deliver now
B5	Named top-options	Is (Brand Name) considered one of the top (Category) options for (Target Audience) in (Region)?	Whether AI places you in the top tier when asked directly

Compare B2 against B1 and B5. If AI speaks well of you when named but picks a competitor when asked openly, you are in a Branded Bubble: visible to buyers who already chose you, invisible to the ones still deciding.

Fan-out queries for this stage

BOFU fan-out gets specific and transactional. AI is trying to confirm details, so the searches look like a checklist:

• (Brand Name) pricing
• (Brand Name) cost for (use case)
• (Brand Name) service area (Region)
• (Brand Name) contract terms
• (Brand Name) how long to get started
• (Brand Name) reviews (current year)
• (Brand Name) vs (main competitor 1) pricing
• best (Category) for (use case) (Region)
• is (Brand Name) worth it

Look at how many of those have one natural answer: a page on your own website. Pricing, service area, contract terms, and onboarding timelines are facts only you can state authoritatively. SurfaceMap shows you which of these each model searched for, and in the verbatim responses you can see where it found an answer and where it gave up and hedged.

Where AI looks for answers

This is where the shift described in the AI Buyer Journey guide becomes decisive.

At TOFU, dozens of third-party pages could answer "best (Category) in (Region)." At MOFU, fewer could answer "X vs Y for (use case)." At BOFU, the buyer's question might carry six constraints at once, and almost no independent source answers all of them. Nobody has published an article about which (Category) provider in (Region) can start within two weeks for (use case) at a mid-market budget.

So AI goes to the source. Typical sources at this stage:

• Your pricing page, or the absence of one
• Your service and use-case pages, for scope and fit
• Your service-area, coverage, or availability information
• Your terms, onboarding, and "how it works" content
• Your FAQs, which often answer exactly the constraint questions buyers ask
• Recent reviews, to sanity-check what your own pages claim

When those pages are clear and specific, AI can confirm the constraints and recommend you with confidence. When they say "contact us for pricing," "solutions for every business," or nothing at all, AI can't confirm, and you'll see the tell-tale hedges in the verbatim response: "pricing isn't publicly listed," "you'd need to confirm availability," "reviews suggest they may be a fit." Meanwhile, a competitor who published those specifics gets the clean recommendation.

At BOFU, your website stops being marketing and becomes the source of record. The question is no longer only whether the information exists. It's whether it exists in a form a machine can find, read, and trust, with the decision-relevant facts stated explicitly instead of buried in sales copy.

What failure looks like

These AI Mention States show up most often as a BOFU problem. See the 12 AI Mention States for the full framework.

• Branded Bubble: strong when named (B1, B5), absent when not (B2).
• Decision Weakling: included at the decision stage but rarely chosen, usually hedged or ranked second.
• Proof Deficit: AI can't find enough evidence (pricing, outcomes, case studies, specifics) to justify recommending you.

How to measure it

• B2 win rate. How often are you the unbranded pick, per model? This is the number that matters most.
• Branded vs. unbranded gap. The difference between B1/B5 and B2 is your Branded Bubble exposure.
• Hedge language. Count the caveats in B1, B3, and B4. Each one points to a fact AI couldn't confirm.
• Constraint confirmation. For each buyer constraint (price, region, timing, use case), did AI find an answer or hedge?
• Who wins instead. When you lose B2, which competitor wins, and which of their pages did AI cite?

How to fix it

On your own site

• Publish the specifics. Pricing or pricing ranges, service areas, onboarding timelines, contract terms, what's included and what isn't. If a buyer would ask it before signing, state it.
• Answer constraint questions directly. Build FAQs around the real BOFU questions buyers ask, one clear answer each.
• Show proof. Case studies with the customer type, use case, region, and outcome clearly stated give AI evidence it can cite.
• Make it machine-readable. Structure decision-relevant facts (offers, pricing models, service areas, inclusions and exclusions, time to value) as structured data, not just prose. The easier the facts are to extract, the more likely AI uses them.

Off your site

• Keep review volume and recency healthy. AI uses recent reviews to check your own claims.
• Make sure third-party listings state the same pricing model, service area, and scope as your site. Contradictions create hedges.

See where you stand at BOFU

• Run a free audit to see whether AI picks you when buyers are ready to decide, and exactly which facts it couldn't confirm when it didn't.
• Take the 2-minute diagnostic to find your AI Mention State.

Previous stage: MOFU · Next stage: EVFU: Do your customers reinforce you, or talk buyers out of you?

FAQ

What is BOFU in AI visibility? BOFU (bottom of funnel) is the Decision stage of the AI buyer journey, where the buyer asks AI which option to choose and AI makes a recommendation against the buyer's specific constraints.

Why does AI recommend my competitor when I'm a better fit? Often because AI could confirm the buyer's constraints (pricing, region, timing, use case) for your competitor but not for you. At the decision stage, independent sources rarely cover those specifics, so AI relies on each brand's own content. The brand that states them clearly usually wins.

What is a Branded Bubble? A Branded Bubble is when AI describes a brand favourably when asked about it by name, but doesn't choose it when a buyer asks an open, unbranded question. It means the brand is visible to buyers who already know it and invisible to those still deciding.

Should I publish my pricing? If buyers ask about price before deciding, AI will try to answer. Publishing pricing, or at least pricing ranges and the pricing model, gives AI something to confirm. Without it, AI tends to hedge or favour a competitor whose pricing it can find.