Voice of Customer Marketing: How B2B SaaS Teams Turn Buyer Language Into Keywords, Prompts and Copy

Voice of customer (VoC) marketing means building your keywords, content and positioning from the exact words buyers use when they describe their problem, compare vendors and explain why they bought. For a B2B SaaS team, that language sits in sales call recordings, support tickets, review sites, community threads and customer interviews. Your job is to collect it, sort it and put it to work on the pages and prompts that drive pipeline.

This work has grown more valuable since buyers moved part of their research into AI assistants. In G2's March 2026 survey of 1,076 B2B software buyers, 51% said they start research in an AI chatbot more often than in Google, up from 29% in G2's 2025 report. Those buyers type long, specific questions in their own words. A team that knows those words can write the pages assistants quote and track the prompts that decide whether the brand gets named.

At Flying Cat, VoC feeds every account. It shapes the keyword map and supplies the prompt sets we track in AI assistants. It also sets the language on money pages.

What voice of customer marketing covers

CX teams use VoC to run satisfaction programs, NPS surveys and churn analysis. Marketing uses the same raw material for a different job.

• Problems in the buyer's words become keywords, headings and prompt themes.
• Trigger events that push a buyer to look for a new tool shape your bottom-funnel pages and your ads.
• Alternatives and objections become comparison pages, FAQ answers and sales enablement.
• Winning outcomes that made them buy become value propositions, CTAs and case study angles.

The distinction affects your tooling. CX platforms like Medallia and Qualtrics are built to score sentiment across thousands of survey responses. A marketing team usually gets further with call recordings, a spreadsheet and a model that can summarize transcripts, as long as a person checks the output.

Where B2B SaaS voice of customer data comes from

Each source captures buyers at a different stage. We pull from at least four of these before writing a strategy.

Sales and call recordings

Sales calls carry heavily commercial language. Prospects ask how you compare with the incumbent, whether you integrate with their stack and what implementation takes. Those are the questions they also ask ChatGPT and Perplexity.

When we rebuilt Bizzabo's content strategy, sales-call insights mapped the ICP's questions, pain points and evaluation triggers. New content followed ICP need and intent strength over raw search volume. Organic visits to Bizzabo's product and solutions pages grew 67% in four months, after non-branded clicks had fallen 23% year over year.

Support tickets

Support data shows the words customers use once they own the product. Those words often differ from the terms on your feature pages. A ticket that says "the export keeps timing out" tells you more about search phrasing than an internal spec that calls the feature "bulk data extraction".

Review sites

Note from Maeva: For SaaS, G2 is my first stop because you can sort reviews by most recent and see each reviewer's role.

Capterra, GetApp, TrustRadius and Software Advice fill in the picture. Read reviews from the past 12 months, since older ones describe a product and market that have moved on.

Review mining works best when you cover competitors too. When we worked with a workforce management client in social care, we mapped each product pillar to its closest competitors and read reviews for every one.

In the Planday review above, ease of use comes through as the main theme. The phrase "couldn't do the job without it" is what copywriters call sticky language. Sticky phrases are short, vivid and easy to remember, so they work in headlines and CTAs.

Forums and communities

Communities show how buyers talk to peers when no vendor is listening. Reddit is the obvious starting point. It also carries extra weight in AI search. Peec's March 2026 analysis of 30 million sources found Reddit ranked first or second on every AI engine it tested, including ChatGPT, Gemini, Perplexity, Google AI Mode and AI Overviews. The threads your buyers read are often the same threads assistants cite.

Niche Slack groups and LinkedIn are just as useful in B2B. Kate Williams, a content lead on our team, joined sales communities like RevGenius when she started writing for Mixmax, a sales engagement platform. "At first, it was difficult to nail the Mixmax TOV," Kate says. "But then, I joined a bunch of Slack groups for salespeople and started paying attention to how they spoke to each other."

She picked up the dry, ironic humor salespeople use with each other and brought it into the blog.

Mixmax also built every SEO article for reuse on LinkedIn, with quotes from its own team and industry experts who then shared the pieces. Blog-assisted demo requests grew 500% in ten months. Keywords in positions 1 to 3 grew 462%.

Customer interviews

Interviews give you the trigger events and emotional context that no dashboard shows. They also surface keywords tools miss. For Softr, customer interviews surfaced "client portal" as the phrase buyers used for what the product did. We then added industry, role and job modifiers such as "construction client portal" and "agency client portal" to the bottom-funnel plan. Softr grew SEO-sourced MRR 227% and SEO-assisted signups 254% in six months.

Voice of customer interview questions for B2B buyers

Althaea Sandover, who wrote the first version of this guide, built the interview framework we still use. It follows the buying journey from goal to outcome. Twenty to thirty minutes with five to eight customers who fit the ICP is usually enough to see patterns.

1. Goals. Find out what the buyer is measured on. - What are your top priorities at work right now? - What does success look like, and how do you measure it?
2. Pains. Find out what they struggled with and what they tried first. - When you first found [product], what problem did you hope it would solve? - What did you use before, and what went wrong with it?
3. Trigger event. Pin down the moment the problem became urgent. - When did you first think you needed a new solution? - What was the breaking point that made you look?
4. Objections. Learn which alternatives they weighed and what nearly stopped them. - What other options did you consider, and why did you rule them out? - Did anything almost stop you from choosing [product]?
5. Outcome. Capture the winning value proposition in their words. - What was the number one reason you chose [product]? - How did you describe it when you recommended it to a colleague?

Follow-up questions produce the best quotes. When the customer sounds frustrated, ask why that bugs them. When they sound pleased, ask why it was a big deal. Ask them to walk you through how the decision got made, including who else signed off.

Record every call and take margin notes as you go. Put themes on one side of the page and follow-up questions on the other, so you can wait for a pause instead of interrupting.

Using AI to analyze calls and transcripts

AI makes VoC research practical at a scale a content team could never reach by hand. A model can read 50 call transcripts in minutes and pull every competitor mention, objection and feature request. That speed is useful as long as you treat the output as a first draft.

Two risks need a human check. Transcription tools sometimes invent words. Cornell researchers found that about 1% of Whisper transcriptions contained entirely hallucinated phrases across more than 13,000 audio clips. Summaries also smooth buyer language into the model's own phrasing, which strips out the sticky words you went looking for.

We follow these rules for AI-assisted VoC analysis.

• Ask for verbatim quotes with timestamps. Tell the model to extract exact sentences, never paraphrases. Have it cite where each one appears.
• Spot-check against the recording. Listen to a sample of quotes in context, especially any you plan to publish or use in positioning.
• Run the extraction by theme. Separate prompts for objections, competitor mentions, trigger events and feature language give cleaner output than one broad summary.
• Keep the tone notes human. Pauses, hesitation and enthusiasm rarely survive a transcript. The person who ran the call adds them.
• Strip personal data before analysis. Remove names, emails and account details from transcripts before they go into any model.

A useful extraction prompt looks like this.

From the transcripts below, list every question the prospect asks about alternatives, integrations, pricing or implementation. Quote each question word for word, give the transcript name and timestamp, and tag it as objection, comparison, feature or process. Do not summarize or rephrase.

Organizing VoC research in a Customer Desire Map

Raw quotes are only useful if writers can find them again. At Flying Cat we keep them in a Google Sheet called the Customer Desire Map (CDM), a format our founder, Maeva Cifuentes, introduced years ago that the team has refined across dozens of clients.

Each quote goes in with its source, the speaker's role and one of four categories.

• Hopes and dreams cover the buyer's goals and what they wish they could do.
• Successes cover what the product has helped them achieve.
• Pains cover their biggest daily challenges.
• Barriers and fears cover what stops them from reaching their goals or switching tools.

We add a themes column for recurring ideas, bold the sticky phrases and keep a swipe copy tab for lines writers can adapt.

The screenshot above is organized by competitor because that client needed alternatives pages. Other accounts sort by persona. When the end user is an engineer and the buyer is a CTO, we keep their quotes on separate tabs, since the two speak very differently.

Turning voice of customer data into keywords, prompts and page copy

The CDM becomes useful when it changes what you publish.

Keywords

Check every recurring phrase in a keyword tool and in Search Console. Some phrases will show search volume you missed, as "client portal" did for Softr. Others will show none. Those often make the best bottom-funnel pages, because the buyers who type them are close to a decision. Trevor.io's plan worked this way, with low-volume alternatives and comparison pages chosen for intent over volume.

In Search Console, filter for queries of ten words or more. Long conversational queries show how buyers phrase the need when they type the way they talk.

AI prompt sets

Prompt research decides what you measure in AI search, so it should rest on evidence. Our prompt sets draw on voice of customer data from sales calls, forums and support tickets, plus Search Console data and some simulated prompts written for specific personas. We group prompts into themes, such as "best tool for a use case" or "replacing a named competitor", and write five to eight prompts per theme. The full method, including a prompt set builder, sits in the prompt research chapter of our AEO guide.

Here is how one call snippet becomes a tracked prompt.

The prompts never include your brand name. A prompt that names you will mention you almost every time, so it tells you nothing about visibility.

Page copy and FAQs

VoC tells you which words to use on money pages and which questions to answer. The FAQs on Flying Cat's own site come from questions buyers asked us on calls, and we build client FAQs the same way.

Maeva's own research supports writing plainly. In a study of 488 product pages from Forbes Cloud 100 companies, she found that descriptive copy correlated with AI citations at 0.19 while benefit-led copy correlated at -0.18. Buyer language tends to be descriptive by nature. Customers say what the product does for them in concrete terms, and assistants quote pages that do the same. Use VoC to describe features and outcomes in the buyer's words, then keep the persuasion for headlines and CTAs. Your overall tone still comes from your brand voice, and VoC supplies the vocabulary inside it.

Measuring the impact of VoC marketing

VoC work should show up in pipeline numbers. We judge it on the same KPIs as the rest of the content program.

• Conversion rate on VoC-led pages. We track visitor to demo or trial conversion rate as a primary KPI, along with SQL to opportunity rate using the client's own definition of a qualified lead. Compare pages rewritten with VoC language against their own baseline.
• AI visibility on VoC-based prompts. Mention rate is the share of answers that name your brand across the prompt set. Report it by theme, since one rolled-up number hides where you are absent. The measurement chapter of our AEO guide covers citations and mention rate in detail.
• Self-reported attribution. Add a "Where did you hear about us?" field to demo and signup forms. AI-sourced pipeline is hard to attribute any other way. The answers also tell you which assistants your buyers use.
• Sales feedback. Ask reps whether prospects arrive using the language on your pages. When a buyer repeats your comparison page back to sales, the research has done its job.

Refresh the research once or twice a year. Run it again after a major launch or a competitor's move. Buyer language shifts as categories mature. AI assistants also change the questions people bring to sales.

How Flying Cat runs VoC research

On every new account, VoC is part of the Strategy Launch, the three to five week phase that opens each program. Maeva leads the strategy herself, with freelance specialists who know the client's vertical. We run the work in this order.

1. Collect. We request 20 to 30 recent sales call recordings, a support ticket export and any interview notes. We add review and community research on the client and its top competitors.
2. Extract. AI pulls verbatim quotes by theme. A strategist checks a sample against the recordings.
3. Organize. Quotes go into the Customer Desire Map by category, persona and competitor.
4. Apply. The CDM feeds the keyword map, the AI prompt set and the briefs for the first money pages. Every brief carries the buyer phrases the writer should use.
5. Measure. We set a baseline for conversions and AI visibility, then report against it monthly.

The research then flows into the content strategy workflow we run from topic approval to quarterly review. If you want buyer language behind your SEO strategy, your content or your AI search program, book a call and we can walk through what your calls and reviews already tell you.

FAQ

What is voice of customer marketing?

Voice of customer marketing uses buyers' own words from calls, reviews, tickets, forums and interviews to shape keywords, content, positioning and campaigns. It differs from CX-led VoC programs, which focus on satisfaction scores and retention.

How is AI used in voice of customer analysis?

AI transcribes calls, extracts quotes by theme and groups recurring objections across large volumes of transcripts and reviews. A person should check quotes against the source, since transcription tools can invent phrases and summaries tend to lose buyer wording.

How often should B2B SaaS teams refresh VoC research?

Once or twice a year works well for a typical B2B SaaS team. Refresh sooner after a major product launch, a pricing change or a new competitor entering the category.

How many customer interviews do you need?

Five to eight interviews with customers who fit your ICP usually reveal the main patterns. Add 20 to 30 sales call recordings and recent reviews to confirm them.