In Forrester's Buyers' Journey Survey of nearly 18,000 global business buyers, twice as many named generative AI or conversational search as a more meaningful research source than any other source, outranking vendor websites, product experts, and sales. The proportion of B2B buyers using AI in their buying process climbed from 89% in 2024 to 94% in 2025. The buying decision is happening inside the AI conversation now, before your sales team ever hears about it.
Visibility tracking tells you whether you appear in that conversation. It cannot tell you whether you win it. Buying simulations can.
Ansehn is the growth platform for the agentic web. We help B2B companies understand how their buyers research, evaluate, and select vendors when using AI, and turn that visibility into revenue. Buying simulations are how we answer the question every marketing team is now asking: when an AI runs the full evaluation, do we win?
Ansehn's buying simulations run the full buyer decision journey through ChatGPT, Gemini, Perplexity, or Claude, with your most relevant persona asking the questions they would actually bring to an AI. The output is a verdict: did you win or lose, against whom, and why. This is the feature walkthrough.
A note on the screenshots in this post: they come from a demo simulation we ran on BMW to illustrate the feature end to end. BMW is not a current Ansehn customer. The data shown is genuine output from a working simulation, but it reflects a sample run, not a customer engagement.
Win rate is the metric that matters now
A brand can be cited in most relevant AI responses and still lose the majority of buying decisions that follow. The AI mentions you in turn one. By turn five, after the buyer probes deeper, it recommends a competitor with stronger reviews or clearer proof. You scored the visibility point. You lost the deal.
Win rate captures the only outcome that matters at the end of an AI conversation: whether the final recommendation is you or someone else. Every section of the buying simulations feature is built around this metric.
Related: Most B2B teams shouldn't track more than 100 prompts. Here's why.
Configure a run: persona, model, market, question
Every simulation starts with four inputs. Pick a buyer persona. Pick the AI search engine to run on. Pick the geographic market. Write the buying question.
Setup for a single run: one persona, one model, one market, one buying question.
The persona is where most teams underinvest. Generic personas produce generic answers. "B2B marketer" tells you almost nothing. "Head of Marketing at a 400-person SaaS company, frustrated that her analytics don't show conversion impact, reporting to a CEO who wants pipeline attribution this quarter" produces a decision journey that mirrors how your most relevant buyer actually behaves. Ansehn ships with a persona library tuned to your ICP, and you can add custom personas as your buyer mix evolves.
The model and market matter for a reason most GEO tools ignore. AI behavior differs across ChatGPT, Gemini, Perplexity, and Claude, and it differs across countries. The same buying question can produce a 22% win rate in Switzerland and a 5% win rate in Germany. If you only test one combination, you are not testing where your buyers are.
Read a single run: the conversation that decided it
The simulation runs as a multi-turn conversation. The buyer asks the initial question. The AI responds. The buyer probes further. By the final turn, the AI recommends a single vendor.
A buying simulation runs as a multi-turn conversation. The fleet manager pushes back on the first answer, asks for independent data, and the AI commits to a final recommendation. In this run, BMW lost to Mercedes-Benz.
This view is where the diagnosis starts. You see the exact turn your brand fell out of consideration. You see which competitor the AI named, and which third-party sources it leaned on (industry blogs, your competitor's own site, review platforms). You see whether the AI cited your content at all, and whether it used that content to recommend you or a competitor.
Every run produces an Analysis Report alongside the conversation: the verdict (Won or Lost), the persona's reasoning, the criteria that decided the outcome, the AI's final recommendation with rationale, and the specific content gaps that caused the loss. See a sample report (PDF) for the full BMW versus Mercedes run shown above. One run is one data point. The pattern only becomes clear when you aggregate.
Diagnose the pattern: the aggregated report
Run a batch of simulations across your main personas and the report turns those journeys into a single diagnosis you can act on.
The aggregated report opens with a win rate, a central finding written in plain language, and the buying journey broken into phases.
The central finding is the part that changes how a marketing team talks. In the BMW simulation above, it reads: "BMW is losing large fleet deals due to a perceived lack of a dedicated, transparent, and digitally integrated fleet program, failing to address buyers' primary concerns around TCO and operational risk." Bring that sentence into a quarterly review and the conversation shifts from "our visibility is up 12%" to "our fleet program looks fragmented to the buyers who matter most." That second conversation gets budget approved.
Below the finding, the buying journey is broken into phases. For fleet buyers, that means Financial and Risk Assessment (TCO, depreciation, reliability), Operational Integration (leasing rates, charging downtime, ecosystem integration), and End-User Experience (driver comfort, insurance costs). Each phase surfaces the criteria buyers actually weigh, in the order they weigh them. You learn where in the journey you lose.
From diagnosis to prescription: content gaps and recommendations
The most actionable part of the report comes next: ranked content gaps and ranked recommendations.
Each content gap is a specific reason buyers chose a competitor. Each recommendation is a specific content action to close it.
This is what makes buying simulations operationally useful. Instead of a vague directive like "improve your messaging," the report names the gap concretely: "BMW lacks a visible, all-in-one charging solution for fleets that bundles hardware, software, installation, and unified billing, creating operational complexity for buyers." That sentence is a brief for your content team. It tells them what to write, who it is for, and why the buyer needs it.
The recommendations work the same way. "Develop and Promote a Branded, Integrated Fleet Charging Ecosystem" is mapped directly to the content gap above it. The simulations are not guessing what buyers want. They are recording it, across many decision journeys, and surfacing the patterns. A content team can take this report into a planning session and walk out with the next quarter's editorial roadmap.
Find the criterion that is deciding your deals
If you want the sharpest single view in the product, open the Decision Criteria Stack Ranking.
Decision criteria ranked by how often they appear in buying journeys, with your win/loss rate on each one.
This is the diagnostic view. In the BMW data above, Fleet Leasing Rates and Charging Ecosystem Integration both show 0% won. Every recorded decision on these two criteria went to a competitor, and they map directly to two of the report's critical content gaps. Fix those, and win rate moves. The same view shows which criteria you are winning on (Maintenance and service costs, at 60% won), so you know where to hold position while you fix the rest.
Track win rate over time
Schedule recurring simulations per persona on a daily, weekly, or monthly cadence and win rate becomes a continuous metric. This matters because a one-off batch gives you a snapshot, not a baseline. When you close a content gap on fleet charging and want to know whether it moved BMW's win rate against Mercedes from 33% to 50% over six weeks, only scheduled runs can answer that. They also catch competitor moves before your pipeline does.
When to use buying simulations vs. prompt monitoring
The two are complementary. Prompt monitoring is your continuous pulse on visibility. Buying simulations are your continuous pulse on win rate, plus the deep diagnostic that explains why.
| Prompt monitoring | Buying simulations | |
|---|---|---|
| What it measures | Whether you appear in AI responses, your rank, share of citations | Whether you win the buying decision against named competitors, for a specific persona |
| What it misses | Why buyers choose competitors. Win rate. Decision criteria. Persona-specific objections. | Per-prompt visibility trends across the broader prompt landscape |
| What you do with it | Track visibility. Spot when you fall out of relevant prompts. | Diagnose why you lose. Prioritize content fixes. Track win rate. Prove ROI. |
| Best used for | Awareness of presence | Decisions on positioning, content, and competitive response |
Prompt monitoring shows you that visibility dropped 8 points this month. Buying simulations dig into why: which competitor took those deals, on what criteria, and what content you'd need to win them back next quarter.
FAQ
How is a buying simulation different from prompt monitoring? Prompt monitoring tracks whether your brand appears in AI responses across a set of tracked prompts. Buying simulations run a full multi-turn buying decision and tell you whether you won against named competitors, for a specific persona, on the criteria that decided the outcome. Use monitoring for breadth and continuity, simulations for depth and root-cause analysis.
How are personas built? Ansehn generates personas from your brand details, market, and competitive context. You can also add custom personas with their own pain points, decision criteria, and buying questions. The quality of the persona determines the quality of the simulation, so investing time here pays off across every run.
What does a "loss" in a simulation actually tell me? Each loss is logged with the winning competitor, the decision criteria that drove the recommendation, and the conversation turns where your brand fell out of consideration. Across many simulations, the losses cluster into patterns. Those patterns are the content gaps you need to close.
How often should I run simulations? Even a small batch (10 to 20 runs) across your main personas can surface clear patterns. Larger batches (50 to 100+) increase confidence in win rate as a metric. For ongoing tracking, schedule recurring runs (daily, weekly, or monthly) for your top personas and key questions. Scheduled runs let you attribute changes in win rate to specific actions.
Who this is for
Buying simulations are most valuable for marketing, PR, growth, and content teams at B2B companies who want deep insight into how buyers actually decide. If your job is to produce content that influences those decisions, win share against named competitors, and tie that work back to pipeline and revenue, this is the view that tells you where to aim.
Run your first simulation
Pick the persona that represents the buyer you most want to win. Write the buying question they would actually ask. Run the simulation. See whether you win, against whom, and why.