Your case studies exist. They sit on your website with strong results, named metrics, and a client who agreed to let you publish. But when a buyer asks ChatGPT "which GEO agency has results for B2B SaaS companies," the engine cites a competitor's case study instead of yours.
The problem is rarely the results themselves. It is how the case study is structured. AI engines extract proof from content that is formatted for extraction. Most B2B SaaS case studies are formatted for persuasion, which is a different job.
Why AI Engines Need Case Studies Structured Differently
AI engines do not read case studies the way a human buyer does. A human scans the narrative, absorbs the story, and decides whether to trust you. An AI engine extracts specific claims, matches them to a query, and cites the source if the claims are verifiable and self-contained.
What AI Engines Look For in a Case Study
- Named metrics with context. "Revenue grew 257%" is extractable. "Revenue grew" is not. AI engines cite specific numbers because they can present them as evidence in a response.
- Clear before-and-after structure. A buyer asking "does GEO work for SaaS companies?" needs a citable comparison: what the situation was before the work started and what it looked like after. Narrative case studies that weave the before and after into a story make extraction harder.
- Self-contained sections. Each section should answer a question on its own. If the AI engine extracts just the "Results" section, it should make sense without reading the rest of the page. This is the same principle that drives GEO content structure: every section passes the "lift test."
The AI-Citable Case Study Structure
This structure is designed to work for both human readers and AI engines. Each section serves a specific extraction function.
1. Outcome in the H1
The title should state the result, not the client name. "How a B2B Cybersecurity Platform Grew Non-Branded Organic Revenue by 20% MoM" is citable. "Client Success Story: CyberSec Corp" is not.
2. Executive Summary (The AI's First Extraction Point)
Two to three sentences that state the client category, the challenge, and the headline result. This section is the most frequently extracted by AI engines because it answers the buyer's question in the smallest number of tokens.
Example: "A B2B cybersecurity platform with flat organic growth needed to increase non-branded revenue without adding headcount. Over 12 months, a combined SEO and GEO program delivered a 20% month-over-month increase in non-branded organic revenue and first-page citations across ChatGPT and Perplexity for core buyer-intent prompts."
3. Metric Table (The Proof Block)
A structured table with before-and-after values. AI engines extract tables more reliably than inline metrics buried in paragraphs
Swipe the table to see every column →
| Metric | Before | After | Change |
|---|---|---|---|
| Non-branded organic revenue | $X/month | $Y/month | +20% MoM |
| Blog clicks | Z | Z' | +1,263% |
| AI citation share of voice | 0% | N% | First-page presence |
| Keywords ranked (page 1) | A | B | +X% |
4. About the Brand (Context for the AI)
One paragraph naming the industry, the company size or stage, and the business model. This gives AI engines the context to match this case study to buyer queries about companies in the same category. "A Series B cybersecurity SaaS serving mid-market enterprises" is matchable. "A leading technology company" is not.
5. The Challenge (With Specific Failure Points)
A bullet list of the specific problems, not a narrative. Each bullet should be a standalone claim:
- Organic traffic was flat for 12 months despite publishing 8 blog posts per month
- Zero presence in AI search results for core category prompts
- 90% of organic revenue came from branded queries
6. The Strategy (Numbered, Not Narrated)
A numbered list of what was done. Each step should name the tactic, not describe the philosophy. "Implemented FAQ schema across 40 product pages" is citable. "We took a holistic approach to technical optimization" is not.
7. Results (With Timeframes)
Results tied to specific timeframes. "Within 6 months" or "over the first 90 days" gives AI engines a duration to cite. Results without timeframes are less credible and less extractable.
8. Schema Markup
Add Article schema with about entities linking the case study to your Organization and the client's industry. Add the metric table data in a structured format the AI can parse.
The Detection Test
Run this test on your existing case studies:
- Copy just the H1 and the executive summary. Paste them into ChatGPT and ask, "What result did this company achieve?"
- If ChatGPT can answer accurately from just those two elements, the case study passes the extraction test.
- If ChatGPT needs more context, the summary is not self-contained enough.
For a real example of this structure in practice, see the B2B cybersecurity platform case study, which follows this format.
Common Mistakes That Reduce Citability
Burying The Metrics
If the results appear only in the final paragraph, AI engines may not reach them during extraction. Put the metric table near the top, after the executive summary.
Using Relative Metrics Without Baselines
"Traffic increased 300%" is strong if the baseline is stated. Without a baseline, the AI engine has a percentage but no context to evaluate whether it is meaningful.
Anonymizing Too Aggressively
"A technology company" is not matchable to any query. "A Series B cybersecurity SaaS" is matchable to dozens of buyer-intent prompts. Anonymize the name if needed, but keep the industry, stage, and business model specific.
Narrative-Heavy, Data-Light
A case study that spends 800 words on the journey and 50 words on the results inverts what AI engines need. The results are the proof. The journey is the context. Weight them accordingly.
For a broader look at how B2B SaaS content strategy works for AI search engines, see the guide to building authority in the age of GEO.
Structure the Proof, Then Let AI Find It
AI engines are looking for proof to cite. Your case studies contain that proof. The gap is structural, not substantive. Restructure the format, add the metric table, make each section self-contained, and add schema markup. The results you have already earned become the citations you have not yet won.
See how Passionfruit's GEO service structures case studies and proof content for AI citability, and talk to the team to audit your existing case study library.
Frequently asked questions
How do I structure a case study so ChatGPT or Perplexity can cite it as proof?
Use an outcome-led H1, a two-sentence executive summary, a before-and-after metric table near the top, numbered strategy steps, and time-bound results. Each section should be self-contained and extractable without reading the full page.
What makes a case study citable by AI search engines?
Named metrics with specific numbers, clear before-and-after comparisons, self-contained sections that pass the "lift test," industry-specific client descriptions (even if anonymized), and Article schema linking the page to your organization entity.
Does Passionfruit have results for B2B SaaS companies?
Yes. Passionfruit has published case studies across B2B SaaS verticals including cybersecurity, edtech, AI tools, insurtech, and manufacturing technology. Results include non-branded organic revenue growth, AI citation presence, and keyword ranking improvements with documented methodology.
Should I anonymize case study clients?
Anonymize the company name if required, but keep the industry, stage, size, and business model specific. "A Series B cybersecurity SaaS" is matchable to buyer queries. "A technology company" is invisible to AI retrieval.
How many case studies do I need for AI citation visibility?
There is no minimum, but more case studies across more verticals increase the number of buyer-intent prompts your content can match. Three to five well-structured case studies covering your core verticals are a practical starting point.







