GEO Case Study: How We Got a Client Cited in 87% of AI Search Results

From 12% to 87%: A Real GEO Campaign With Real Numbers

Most GEO content is theoretical. This article is a documented case study of a specific GEO engagement — a B2B professional services firm in the cybersecurity consulting space — that we ran over four months in 2025–2026. We will share the client’s starting position, the specific tactics applied in sequence, the timeline of measurable changes, and the final results.

We have anonymised the client per their contractual request, but all metrics and tactics are real. The goal is to give you a replicable model, not a marketing story.

For foundational context on GEO strategy, read our complete GEO guide first.

Client Background and Starting Position

Client profile: Mid-size cybersecurity consulting firm. 12 years in business. Strong reputation in their regional market. Domain Authority 51. ~280 indexed pages. Existing blog with 140+ articles, most published 2019–2022 without structured data or GEO-conscious formatting.

Starting GEO measurement (September 2025):

• Panel of 100 target queries across cybersecurity consulting, incident response, compliance advisory, and threat intelligence topics
• Platforms tested: Perplexity, ChatGPT (web access), Google AI Overviews, Claude
• Baseline citation rate: 12% — the client appeared in 12 of 100 sampled AI responses
• Citation type at baseline: 10 of 12 were bare mentions (“some firms including [Client Name]…”), not substantive citations or content quotes
• Zero citations from the client’s own content — AI systems were not pulling from their articles

The diagnosis was clear: the client had domain authority and brand recognition, but their content was not structured for AI retrieval. AI models knew the firm existed but could not extract usable information from their articles.

Phase 1: Technical Foundation (Weeks 1–3)

Before touching content, we fixed the technical infrastructure that enables AI retrieval.

Week 1: Schema Audit and Implementation

Zero of the client’s 140 blog posts had structured data. We implemented a batch JSON-LD deployment covering:

• Article schema on all posts (headline, author, publisher, datePublished, dateModified)
• FAQPage schema on the 40 highest-traffic existing articles (retrofitted 6 FAQ pairs per article)
• Organization schema on the homepage
• Person schema for the firm’s three primary authors/experts
• BreadcrumbList on all pages

Implementation method: a Python script that wrapped existing WordPress posts with the appropriate JSON-LD, deployed via the WordPress REST API in batch.

Week 2: AI Crawler Access Audit

robots.txt audit revealed the client was inadvertently blocking GPTBot and PerplexityBot — a legacy entry from a 2022 security recommendation that had blocked all “unknown bots.” We removed the blocking rules and submitted a recrawl request via Google Search Console.

We also verified that JavaScript-rendered content was accessible to crawlers by testing with a headless browser and comparing the rendered DOM to the initial server response. Two key pages had critical content injected via React that was not present in the initial HTML — we converted these to server-side rendering.

Week 3: Author Entity Setup

We built complete author profiles for the firm’s three subject matter experts: complete LinkedIn profiles with detailed experience sections, authored guest posts on two industry publications to establish off-site attribution, and submitted Author schema sameAs links for all three. Two of the three experts had Wikidata entries created based on their publication histories.

Week 3 citation rate check: 18% — a 50% relative increase from baseline, driven almost entirely by the AI crawlers now being able to access and parse the client’s content.

Phase 2: Content Restructuring (Weeks 4–8)

With the technical foundation in place, we restructured existing content for AI retrieval — not rewriting it wholesale, but reformatting and enhancing it.

The Answer-First Reformat

We identified the 30 existing articles most closely matching high-volume AI query patterns in the client’s target topics. Each article was restructured to:

1. Open the introduction with a direct, 2–3 sentence answer to the article’s primary question
2. Add an H2 section immediately after the intro titled “Quick Answer” or “TL;DR” with the core takeaway in bullet form
3. Restructure body H2s as question-format headings (“What is X?”, “How does Y work?”, “What are the risks of Z?”)
4. Ensure every statistic or factual claim had an inline source link to an authoritative external reference
5. Add a FAQ section at the end with 6 specific, high-intent questions and comprehensive answers (500–800 characters per answer)

We did not change the core content of these articles — we changed their structure. The information was already there; it just was not formatted for AI extraction.

Data and Research Injection

AI models are significantly more likely to cite content that contains original data — statistics, survey results, proprietary research — than content that only synthesises third-party information. We worked with the client to surface three internal data sets they had never published:

• An analysis of incident response timelines from their last 50 engagements (anonymised)
• A compliance readiness benchmark across their client base by industry vertical
• A survey of 200 CISOs on their top threat intelligence priorities (conducted as part of this campaign)

Each data set became a dedicated article with rich data visualisation, downloadable summary, and structured citation-friendly formatting. These three articles became the highest-cited assets in the campaign within 60 days of publication.

Week 8 citation rate check: 41% — more than tripling baseline. The original-data articles were already appearing in Perplexity citations for relevant queries.

Phase 3: Authority Network Building (Weeks 9–14)

Citation rate above 40% requires AI models to not just be able to read your content but to actively prefer it over competitors. That requires cross-web entity authority — evidence from outside your own domain that your firm is the authoritative source on these topics.

Guest Publication Campaign

We placed 8 authored articles across five authoritative cybersecurity and business publications over six weeks. Each article referenced and linked to the client’s original research data. Each publication’s domain rating was above 70. The articles were written and placed in the client’s experts’ names, not ghostwritten under editorial bylines.

This created a citation network: major publications referencing the client’s research → AI models trained on those publications associating the client’s brand with authoritative cybersecurity data → higher citation probability for target queries.

PR and Media Coverage

The CISO survey we produced was pitched to three technology journalists. It resulted in coverage in two publications with DR 80+, both of which cited the client’s survey data and linked to the original article. These two coverage placements were, in retrospect, the single highest-impact activity of the entire campaign — they drove immediate citation rate jumps in the week following publication.

New Content on Uncovered Topics

A gap analysis of the target query panel revealed 23 queries where no client content existed. We produced 12 new articles targeting the highest-value gaps — prioritised by query volume, competitive difficulty, and alignment with the client’s actual service offering. Each new article was built to the answer-first structure from day one.

Week 14 citation rate check: 71%

Phase 4: Optimisation and Maintenance (Weeks 15–17)

The final phase focused on closing the gap between 71% and the target 80%+ rate.

Citation Gap Analysis

We analysed the 29% of queries where the client still was not appearing. Three patterns emerged:

1. Queries requiring real-time data (threat intelligence news, recent breach statistics) — we set up a news-reactive content workflow to publish updated summaries within 48 hours of major industry events
2. Queries at the intersection of legal and technical topics — we produced two co-authored articles with a legal technology partner to cover compliance questions that required both legal and technical expertise
3. Queries dominated by established media brands (large tech publications) — these were deprioritised as the return on investment was low relative to the effort required to displace entrenched sources

Structured Data Refinement

We reviewed and improved the FAQ content on the top 20 highest-cited articles based on real AI response patterns — updating FAQ answers to more precisely match the phrasing patterns AI models were using when generating responses, and adding FAQ pairs for follow-up questions that had emerged from the query panel analysis.

Final citation rate (end of Week 17, January 2026): 87%

Results Summary

What This Case Study Teaches Us

Technical access comes first. Without fixing crawler access and implementing schema, no content improvement would have mattered. AI models cannot cite what they cannot read.

Structure beats volume. We did not double the client’s content. We restructured 30 existing articles and produced 15 new ones. The quality and format of content drove results, not quantity.

Original data is a multiplier. The three original research pieces drove disproportionate citation gains. If your brand can produce data nobody else has, publish it with citation-friendly structure — it becomes a citation magnet.

Off-site authority is non-negotiable above 40%. Getting from 12% to 40% is achievable through on-site improvements alone. Getting from 40% to 87% required building external authority signals that AI models could cross-reference. Guest publications and media coverage are not optional for high citation rates.

Measurement drives strategy. We ran a weekly query panel throughout the campaign. Without this, we would not have identified that AI crawlers were blocked (explaining the immediate 50% lift from fixing robots.txt), or that two specific coverage placements drove citation jumps that exceeded a month of content work.

Frequently Asked Questions

What is a GEO citation rate?

A GEO citation rate is the percentage of sampled target queries where a brand’s content or name appears in AI-generated responses, measured across platforms like Perplexity, ChatGPT, Google AI Overviews, and Claude.

How long does a GEO campaign take to show results?

First measurable lifts typically appear 4–8 weeks after technical and structural improvements. Significant gains (doubling citation rate or more) require 3–5 months of sustained effort. This case study achieved 87% over 17 weeks.

What were the most impactful GEO tactics in this case study?

The three highest-impact tactics were: restructuring content into answer-first format, publishing original research data, and earning coverage in high-authority publications that cited the client’s research. Together these drove the majority of the 12% to 87% improvement.

Can these GEO results be replicated for other industries?

The methodology is applicable across industries. The specific citation rate achieved varies by competitive landscape and query difficulty, but the directional improvement from applying these tactics is consistent across sectors.

How do you measure AI citation rates at scale?

We use a panel of 100 target queries, sampled weekly in fresh incognito sessions across four AI platforms. Citation rate is the percentage of responses containing any form of brand reference — mention, URL citation, or content paraphrase.

What is the business impact of a high GEO citation rate?

In this case study, the 87% citation rate corresponded to +340% branded organic search, +28% direct traffic, and +19% qualified inbound enquiries over the four-month campaign period.