You're generating mentions, earning coverage, and building what looks like a healthy third-party presence, and yet AI search is still citing your competitors instead of you. Your PR agency's monthly report is full of green checkmarks: the placements are real, the coverage is legitimate, and somehow, when a potential buyer asks an AI to recommend vendors in your category, your brand doesn't come up.
The gap between those two realities is a sourcing problem.
The short answer: AI rebuilds which sources it trusts around the subject of each query. Generic PR coverage doesn't build AI citation authority. Topic-specific coverage in the sources AI already trusts for your category does. Three targeted placements in the right source move the needle more than twelve scattered across the wrong ones.
Kevin Indig and Amanda Johnson's citation research, published in June 2026, shows that the source mix AI relies on shifts dramatically by topic. Ask an AI about invoicing, and it leans on one set of sources. Ask about starting a business, and it leans on a mostly different one. Same model, same query interface, entirely different source ecosystems. Your PR strategy has to account for that gap, because most strategies currently don't.
The finding that should reshape how every PR budget is allocated: competitor and product review sites account for 33.5% of AI citations for invoicing queries, but only 7% for starting-a-business queries. Research publications and how-to guides flip in the opposite direction. Same AI model, same users, entirely different source sets depending on the question.
This means "authoritative publication" is not a fixed designation. Authority is topic-conditional. The Wall Street Journal carries general credibility, but if AI isn't using it as a primary source for your category, a placement there does very little for your AI citation rate. The sources AI cites for payroll questions are largely different from those it cites for hiring questions, which in turn differ from those it cites for business formation. Every sub-topic has its own trusted ecosystem.
Most PR teams aren't operating at this level of specificity. They're targeting publications by domain authority or general brand recognition, which produces coverage that looks good in a monthly report but doesn't map to the source sets AI is actually reaching for. The result is a mismatch between spend and return that's easy to miss when you're measuring mentions instead of citations.
The fix starts with research, not outreach. Before a single pitch goes out, the question has to be: what sources does AI cite when someone asks about our topic? The answer isn't obvious, and it varies more than most people expect.
The second finding from Indig and Johnson's research is equally important: authority doesn't compound gradually across all sources. It accumulates in tiers, and the movement from one tier to the next is what changes citation rates.
Think of it as a pyramid. The top decile of sources for your topic carries disproportionate weight. Getting three placements there moves AI citations more than a dozen scattered across mid-tier or lower sources. The middle tier is crowded, and incremental additions to it produce incremental results, often too small to measure. The bottom tier produces almost nothing.
This is the opposite of how most PR programs operate. The conventional logic is to maximize coverage breadth, spreading budget across many publications to build as wide a presence as possible. That logic made sense in a world where Google was counting link signals across a broad distribution. When the goal is AI citation, the calculus flips: AI weights a relatively small set of trusted sources for your specific topic and largely ignores the rest.
The strategic implication is to concentrate. Identify the top-tier sources for your category and spend there first and repeatedly until you have meaningful depth. Two or three authoritative placements in the right source build more AI visibility than ten placements in sources that don't appear in your topic's citation ecosystem.
This is also why PR agencies that optimize for quantity of coverage are misaligned with what AI search actually rewards. What matters is whether your brand appears in the specific places AI already trusts for your topic, not simply how many places it appears overall.
AI systems anchor authority to entities. A human author with a recognizable track record, an active publishing history, and visible credentials gives AI something concrete to attach trust to. A brand account without an identified author doesn't give it nearly as much.
When a subject matter expert publishes under their own name, they bring personal authority to the content that a brand byline can't replicate. Semrush's analysis of 89,000 LinkedIn URLs cited in AI search found that the most-cited authors post consistently and have clearly visible credentials, and that active, credible authorship outperformed anonymous or loosely attributed content on every major AI platform. Expert-authored content published under a real name with a clear timestamp earns AI visibility more quickly than brand-attributed content covering the same ground.
The practical move is to identify two or three internal experts who have genuine credibility in your topic area. They don't need to be executives or founders, but they need to know the subject deeply, understand how it connects to your product or category, and be willing to publish consistently. Give them a clear format to work in (how-to guides and original research account for a significant share of AI-cited content) and build their presence as named authors in the sources your topic's AI ecosystem trusts.
This is how earned media has always worked at its best. The difference now is that the authority signal runs directly through the author entity, not just the publication, and AI is paying attention.
Before mapping sources or pitching anything, identify two or three people who can publish credibly on your topic. Look for genuine subject knowledge, a willingness to publish, and some existing public presence. Document their credentials, set up their profiles on major platforms, and establish them as named authors with bylines. The expert entity is the foundation on which everything else builds.
Run your highest-intent queries through three or more AI platforms: ChatGPT, Perplexity, Gemini, and any others relevant to your audience. Record every source that appears. Repeat across 10 to 15 prompts covering the subtopics within your category. Note which domains appear consistently across multiple platforms and which are unique to one. Consistent cross-platform overlap is your highest-priority target list, but don't discard single-engine sources entirely, since research shows that the vast majority of AI citations appear in only one engine, so platform-specific authority still matters.
One tactical note: for AI citation purposes, the distinction between dofollow and nofollow links is much smaller than it is for traditional SEO. The correlation between nofollow links and AI mentions is nearly identical to the correlation for dofollow links, which means nofollow-heavy sources like industry directories, content aggregators, and roundup publications are worth including in your target list even if a traditional SEO audit would deprioritize them. Don't let link-building logic filter out sources that AI actually cites.
Once you've mapped the trusted sources, getting into those publications is only part of the goal. The more important move is getting your expert into the same conversations the trusted sources are already having. That means pursuing quotes from the journalists who appear most in AI responses, co-authoring with writers AI already cites, booking appearances on the same podcast series where recognized voices publish, and commenting substantively on discussions the trusted sources start. Co-occurrence with a trusted entity pulls your expert into the candidate set faster than a standalone post in the same publication, because AI learns authority partly through association.
Rank your target sources by topic-specific authority, not just domain authority. The majority of your PR effort should go toward the top three to five sources in your category, a smaller share toward the next tier, and a monitoring-level investment in emerging sources worth tracking. The exact ratio matters less than the principle: depth in the right places compounds in ways that breadth across the wrong ones simply doesn't.
Original research, charts, and data visualizations published under your expert's byline earn citations across publications you never pitched. One well-executed original dataset can generate attribution across dozens of pages. Format matters here: AI leans heavily on answer-ready content like step-by-step guides, comparison charts, and quantified findings. Publish in those formats, make the content easy to embed and share, and seed it to your top-tier target sources.
Named-author posts on LinkedIn can appear in AI answers within days of publication, particularly when the author already has an established audience and the post earns engagement. OtterlyAI's analysis of 1.3 million LinkedIn citations found that consistent publishing frequency and verifiable credentials were the strongest predictors of AI citation across all six major platforms. This is faster than waiting for a new placement to get indexed and weighted by most AI systems. Publishing under your expert's name, with links back to original research and company resources, builds both the personal-entity signal and brand association simultaneously.
Most PR measurement systems reward the wrong outcomes. A "47 mentions this month" report says nothing about whether those mentions moved AI citation rate, which is the actual goal.
The metrics worth tracking are more specific:
This is the most direct signal that your authority-building is working. Track when each piece of expert content publishes, then query across AI platforms weekly to find when it first appears in a response. Fast integration means the authority signals are landing. Slow or absent integration is a sign the source or the content format isn't resonating with AI's retrieval logic.
Instead of total mention count, track how many of those mentions appear in the specific sources your topic's AI citation ecosystem recognizes. This number starts small and should grow deliberately.
Whether AI cites you once from a given source matters less than whether it cites you repeatedly. Recurring citations signal that AI has integrated that source and your association with it into its model.
Query your expert's name alongside topic keywords across platforms. Track when and how they appear, which platforms surface them first, and what topics they're most often cited on.
Tracking these metrics requires manual querying or tools built specifically for AI citation monitoring. The measurement is imperfect, but it's more signal-rich than counting total mentions across publications that may not appear in AI responses at all. Tools like Profound, Otterly AI, and Peec AI are built specifically for this.
For years, "build authority" was advice that could mean almost anything, and the generic version of that advice (build links, earn press, get coverage) still circulates widely as strategy. The more specific version, the one that actually maps to how AI search works, is to build authority in the sources AI trusts for your specific topic, at a depth that places you in the top tier of that category's ecosystem. That's the first layer of what we call the Architecture of Authority.
That specificity changes where you invest PR budget, who carries the content, what success looks like, and how you measure it. It also reframes the relationship between earned media and organic search performance. Most strategies haven't fully accounted for this yet, because they were designed for a search environment that no longer exists.
At Arcalea, this is the foundation of our Brand Consensus process. We start by mapping the source ecosystem AI cites for a client's category before any outreach begins, because pitching without that map means spending budget on coverage that doesn't reach the systems that matter. If you want to understand where your brand currently stands in AI responses, and which sources you'd need to be in to change that, let's talk.