AI Communications Case Studies: What Changed After 90 Days

Citation Share moves in 90 days. Not because the work is easy — because the infrastructure compounds faster than most companies expect when earned media, GEO, and AI-visibility measurement operate as one system. I've watched it happen across sectors — beauty, technology, financial services, consumer brands, B2B enterprise. Here's what actually changes when a company engages an AI Communications firm and commits to the discipline.

The pattern — what happens in the first 90 days

Every engagement starts the same way: measure the baseline, identify the gaps, build the infrastructure, and let the corpus compound. The specifics vary by sector and starting position — but the pattern is consistent enough to document.

Days 1–30: The baseline reveals the gap

The first thing we measure is Citation Share — the percentage of AI-generated answers to the client's category buyer prompts that mention their brand. We run thousands of prompts across ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews.

The gap is almost always wider than the client expects. Companies that dominate their category in traditional media often own 5–15% Citation Share at intake. Their competitors — sometimes smaller, sometimes newer, sometimes funded — own 30%, 40%, or more. The AI engines don't weight the brand the way the market does. They weight the corpus.

The entity audit reveals why. Inconsistent naming across platforms. Missing or outdated structured data. Wikipedia entries that haven't been updated in three years. Crunchbase profiles with the wrong category tags. Press coverage that the engines can't reliably attribute to the brand because the entity infrastructure is fragmented.

The earned media audit maps source diversity. Most brands — even those with strong press programs — have coverage concentrated in two or three outlet categories. Research from Everything-PR documents that the engines weight source diversity more heavily than source volume. Concentrated coverage doesn't compound the way diversified coverage does.

Days 30–60: Infrastructure builds while placements land

Two tracks run simultaneously.

Track one — earned media. Placements begin landing in tier-one outlets. The pitching strategy is different from traditional PR: we target placements that create high-authority retrieval anchors — specific, quotable, data-rich content that the engines will grab when assembling answers. A quote that says "we're excited about the future" is useless for retrieval. A quote that says "Citation Share in our category shifted 12 points in Q2" is an anchor the engines cite.

Track two — GEO infrastructure. Entity corrections across every platform. Schema deployment — organization, product, FAQ, article. Retrieval anchor engineering on owned properties. Source diversity build-out in the outlet categories where the brand is absent. This is the structural work that most PR firms don't do — and it's the work that determines whether the earned media placements actually compound inside the corpus or just generate clips.

Days 60–90: The corpus starts compounding

By day 60, the first cycle of earned media has been published and indexed. The GEO infrastructure is live. The entity is consistent across platforms. And the measurement layer is tracking.

The 90-day measurement shows the first movement. Not always dramatic — but measurable. Citation Share moves from the baseline by 3–8 percentage points for most mid-market companies. Enterprise programs with larger starting corpora sometimes move faster because they have more existing authority for the infrastructure to amplify.

The more important signal at 90 days is the prompt-level shift. Specific buyer prompts that returned zero brand mentions at intake now return the brand in the answer. "Best [category] firms" starts including the client's name. "Who handles [specialty] for [industry]" starts citing the client's work. These are the prompts that map directly to pipeline — and the client can see exactly which ones moved and which ones didn't.

Composite results — what we've seen across sectors

Beauty & fashion brand — from invisible to cited

A prestige beauty brand with strong retail distribution and solid traditional media presence. Starting Citation Share: 8% for category buyer prompts. Their competitor — a digitally native brand with half the revenue — owned 34%.

The problem wasn't media coverage. They had plenty. The problem was entity fragmentation — five different name variations across platforms, outdated schema, and press coverage the engines couldn't reliably attribute. The competitor had cleaner infrastructure and more diversified source coverage, despite fewer total placements.

After 90 days of AI Communications work — entity consolidation, GEO infrastructure, and earned media targeted at missing outlet categories — Citation Share moved to 19%. At six months: 31%. The competitor's share held steady. The gap closed because the infrastructure compounded — not because we outspent them on media.

B2B technology company — owning the category prompt

A mid-market SaaS company in a competitive vertical. They ranked well on Google. They had a strong content marketing program. Starting Citation Share: 4%. Their three main competitors collectively owned 67% of the AI-engine answers.

The SEO content that ranked on Google wasn't being retrieved by the AI engines. Different signals. Different weighting. The GEO audit revealed that their content was optimized for keyword density — which Google rewards — but lacked the entity-rich, source-diverse authority signals the AI engines use for retrieval.

The 90-day program rebuilt the content architecture around retrieval anchors, placed earned media in six outlet categories where they had zero presence, and deployed full schema infrastructure. Citation Share at 90 days: 14%. At six months: 28%. The lead competitor dropped from 31% to 24% — not because they got worse, but because the engines redistributed citation when a new authoritative source entered the corpus.

Financial services firm — crisis-driven engagement

A financial services company engaged after a negative press cycle. The AI engines were surfacing the crisis coverage in every answer related to the firm — "Is [company] trustworthy," "[company] reviews," "[company] vs. [competitor]." Citation sentiment was running 72% negative across all engines.

Traditional crisis response would have focused on pushing negative results down Google's rankings. AI-engine crisis response is different — you can't bury a result. You have to engineer the corpus so the engines cite the resolution, the correction, the updated narrative alongside (and eventually instead of) the crisis coverage.

The 90-day program placed 14 tier-one earned media pieces focused on the corrective narrative, rebuilt entity infrastructure to associate the brand with the updated positioning, and deployed GEO-optimized content across owned properties. Citation sentiment shifted from 72% negative to 41% negative at 90 days. At six months: 23% negative, with the majority of answers now citing the corrective narrative first.

Consumer brand — launch into a new category

A consumer brand entering a new product category. No existing AI-engine presence in the new category — 0% Citation Share. Five established competitors owned the answer.

Launching into a category where the AI engines already have established answer patterns is harder than it was with Google — because the engines don't just index your content, they decide whether to cite it based on authority signals that take time to build. You can't rank your way into the answer with keywords. You have to earn your way in with authority.

The 90-day program built the category association from scratch: earned media in outlets the engines trust for that category, entity infrastructure linking the brand to the new category, and GEO content architecture creating retrieval anchors for the buyer prompts that defined the category. Citation Share at 90 days: 7%. Not dominant — but present. The brand now appeared in the answer. At six months: 18%. At twelve months: 29% — second in the category, up from zero.

What doesn't work — patterns we've seen fail

GEO without earned media. Structured content and schema alone don't build authority. The engines still weight tier-one publications as highest-trust sources. GEO amplifies earned media — it doesn't replace it.

Earned media without GEO. Press placements that aren't engineered for retrieval generate clips but don't compound inside the corpus. The engines index the coverage but don't reliably cite it because the entity infrastructure isn't clean enough for attribution.

Campaign-based engagement. A 90-day sprint that stops at day 91 loses most of its gains within two quarters. The corpus needs continuous feeding — new placements, updated content, entity maintenance. Stopping a retained program is like stopping an investment program: the compounding stops.

Impressions as the success metric. Every failed engagement we've seen post-mortem had one thing in common: the governing metric was impressions or clip counts, not Citation Share. Impressions measure noise. Citation Share measures whether the AI engines include your brand when a buyer asks the question. They're not correlated the way people assume.

Frequently asked

Are these real client results?

These are composite illustrations drawn from patterns across 5W's client engagements. Specific client names and exact figures are confidential. The patterns — the starting gaps, the infrastructure problems, the 90-day movement ranges, the compounding curves — are representative of what we see consistently across sectors.

What happens if I stop after 90 days?

The corpus starts decaying. Not immediately — but within two quarters, competitors' new content begins diluting your Citation Share gains. The infrastructure doesn't disappear, but it stops compounding. The brands that sustain multi-year programs build structural advantages that new entrants can't replicate in a quarter. Stopping resets the compounding clock.

How fast can Citation Share move for my brand?

It depends on three factors: your starting authority (existing media corpus and entity infrastructure), your category's competitive density, and the source diversity of your current coverage. Mid-market brands with moderate existing authority typically see 3–8 percentage points of Citation Share movement in the first 90 days. The compounding curve steepens in months 4–12. Read the full operating model in The 5W AI Communications Model.

Ronn Torossian is the founder and chairman of 5W AI Communications, the AI Communications Firm. He is the publisher of Everything-PR and the author of two best-selling editions of For Immediate Release.