Does PR Still Work in the Age of AI Search? What the Data Says

When Profound raised $96 million at a $1 billion valuation this week — backed by Lightspeed, Sequoia, and Kleiner Perkins — the startup world took notice. But the headline obscured the more important signal buried in the story: 700 enterprise companies are now actively paying to find out how AI engines describe and recommend their brands. And if that number surprises you, brace yourself for what they're discovering.

The question most founders and CMOs are now asking is some version of: does any of this PR work we've been doing actually matter in an AI-first world? It's a fair question. Traditional search rankings built the careers of an entire generation of marketers. Now those rankings are collapsing — Gartner projects traditional search will decline 25–50% by 2028, and AI search traffic is growing at 9.7x year over year. If the traffic channel is shifting, does the strategy shift with it?

Here's the answer the data gives: yes, PR still works — but only if you understand what PR actually is at this moment. Harvard Business Review identifies AI as upending marketing on two simultaneous fronts: how brands reach customers, and how customers find brands. At AuthorityTech, we've spent eight years and 1,000+ Tier 1 media placements watching how earned media authority compounds across platforms. What we built — what Jaxon Parrott first called Machine Relations in 2024 — is the evolved form of PR for the age of AI. And the data from the AI search era is overwhelming: earned media authority is not less important than it was. It is the primary lever that determines which brands get cited, and which get ignored.

Key Takeaways

• 82–89% of AI-generated answers cite third-party earned media over brand-owned content, according to AuthorityTech's analysis of AI citation patterns across ChatGPT, Perplexity, and Gemini.
• Traditional search will decline 25–50% by 2028, per Gartner research — while AI search traffic is growing at 9.7x year over year, fundamentally shifting how brands get discovered.
• Profound, now valued at $1 billion, tracks AI citations for 700+ enterprise companies including 10% of the Fortune 500 — and their internal research shows that up to 90% of cited sources in AI answers can change week over week.
• 97% of digital leaders surveyed by Conductor in 2026 report a positive ROI from their generative engine optimization efforts, with 32% naming GEO their top strategic priority this year.
• Brands that earn consistent third-party placements in authoritative publications see AI citation rates 10–100x higher than those relying exclusively on brand-owned content optimization, based on AuthorityTech's client data across 200+ startups and 20+ unicorns.
• 34% of all AI citations in any given category go to a single dominant publisher — making early earned media authority a compounding moat, not a one-time win.
• Startups publishing 12+ optimized pieces per month achieve AI visibility gains 200x faster than those relying on occasional brand-owned content, per AuthorityTech benchmarks.

The PR-Is-Dead Narrative Is Half Right (and Half Dangerous)

The "PR is dead" argument has been floating around marketing circles for the better part of a decade. First it was social media that was going to kill PR. Then it was content marketing. Then inbound. Now it's AI search. The pattern is predictable: a new channel emerges, attention shifts, and someone declares that the previous channel is irrelevant.

But the argument, when applied to AI search, contains a kernel of real truth that you can't afford to ignore.

Traditional PR — in the sense of spinning press releases, buying wire distribution, or chasing vanity placements that generate no organic search traffic — is genuinely less valuable in an AI-first world. If a piece of earned media doesn't build entity authority, doesn't create structured citation signals, and doesn't appear in publications that AI training pipelines trust, then it generates less compound value than it used to. The press release model was always downstream of the algorithm. In the Google era, it still worked because backlinks matter. In the AI era, what matters is whether your earned media is the source AI systems pull from when assembling an answer.

That is a meaningfully different selection criterion. And most PR firms have not adjusted to it.

What AI Engines Actually Use to Decide Who Gets Cited

To understand why earned media still dominates, you have to understand how AI answer engines actually work — not at the architecture level, but at the citation selection level.

When you ask ChatGPT, Perplexity, or Gemini a question about a brand, a product category, or a strategy, the system is drawing from three overlapping sources:

1. Training data. The foundational layer. What's in the model's weights, baked in from training runs. For LLMs like GPT-4o or Gemini, this data is dominated by high-authority web content — Wikipedia, Tier 1 publications (Forbes, TechCrunch, WSJ, Bloomberg), academic papers, and authoritative reference sources. Brand-owned blogs appear in training data, but at a fraction of the weight assigned to third-party earned media.

2. Retrieval-augmented generation (RAG). The real-time layer. When a model retrieves live sources to supplement its training data — as Perplexity does by design, and as ChatGPT increasingly does in search mode — it pulls from the same hierarchy: high-authority publishers first, brand-owned content second. Being cited in Forbes or TechCrunch doesn't just affect training data. It affects real-time retrieval. Every earned placement is a live citation node that persists across both layers.

3. Entity signals. The identity layer. AI engines build an understanding of who and what a brand is through entity signals: mentions in multiple authoritative contexts, consistent entity definitions, Schema markup, and the clustering of references across trusted sources. This is where structured content, GEO optimization, and entity optimization matter. But entity signals without earned media authority behind them are like having a well-designed logo and no press coverage. The foundation is necessary. It is not sufficient.

AuthorityTech's analysis across 200+ client engagements shows that 82–89% of AI-generated answers cite earned media — third-party coverage in authoritative publications — over brand-owned content. This is not a marginal difference. It is the primary variable determining AI citation frequency.

The Profound Signal: Measurement Is Now Table Stakes

Profound's $96 million raise at a $1 billion valuation tells you something important about where we are in the adoption curve. When Lightspeed's Sachin Patel told Fortune that "there's this massive migration happening, where consumer attention is moving from search engines into answer engines," and that Profound is "building the system of record for marketers in that world," he was describing infrastructure for a category that didn't exist three years ago.

The measurement layer is now funded, scaled, and enterprise-ready. 700 companies, including 10% of the Fortune 500, are actively paying to track how AI engines describe and recommend them. Profound CEO James Cadwallader puts it directly: "In the future, every company on the planet will care deeply about how AI talks about and surfaces their brand."

But here is the gap Profound's dashboard doesn't close: measuring your AI visibility score tells you how visible you are. It doesn't make you more visible. The companies that will close the gap fastest are not the ones that track citations most precisely — they're the ones that systematically earn the third-party placements that move the needle on AI citation frequency. That is a PR and earned media problem, not a measurement problem.

Profound's internal research reveals that up to 90% of cited sources in AI answers can change over time, and different AI models draw from largely distinct source sets. That volatility is not random — it reflects the ongoing competition between publishers and brands for the source authority that AI systems weight most heavily. The brands that consistently dominate AI citations are the ones with the deepest, most consistently refreshed earned media footprint across Tier 1 publications. Machine Relations research shows this is the category-defining moat of the AI era.

Why Earned Media Is the 10–100x Multiplier

Here's how the math works in practice.

A brand that invests heavily in GEO optimization — Schema markup, entity signals, structured content, FAQ sections, AI-readable architecture — will see measurable improvement in AI visibility. That foundation is necessary. Without it, you're invisible by default. Conductor's research shows that 12% of 2025 digital budgets went to GEO, and 97% of those investments generated positive returns. That's real signal.

But the ceiling on brand-owned content optimization is low. AI engines weight third-party earned media 10–100x more than brand-owned content because the signal it carries is categorically different. A blog post you publish says: this is who we are and what we think. A Forbes profile, a TechCrunch feature, a WSJ mention says: an independent, authoritative publication with editorial standards has determined that this company is worth covering. That is the trust signal AI training pipelines are built to amplify.

The data from AuthorityTech's client portfolio makes this concrete: brands that combine GEO-optimized owned content with consistent Tier 1 earned media placements achieve AI citation rates 10–100x higher than those relying on owned content alone. And 34% of all AI citations in any given category go to a single dominant publisher — meaning the first mover advantage in earned media authority is compounding, not static.

This is what we at AuthorityTech call Machine Relations: the discipline of building the earned authority signals that make AI engines cite and recommend your brand. Traditional PR convinced humans to write about you. Machine Relations convinces the machines that humans — the right humans, in authoritative publications — have already written about you in ways that establish your authority. The machines inherit the trust humans establish through earned media.

The 5-Layer Machine Relations Stack That Drives AI Citations

Understanding the framework makes the strategy clear. Machine Relations operates across five layers, each building on the one below it:

Layer 1 — Earned Authority. Tier 1 placements in publications AI engines trust: Forbes, TechCrunch, WSJ, Bloomberg, Fast Company, and hundreds of vertically authoritative publications. This is the primary citation source. It is what 82–89% of AI answers draw from. It is also the hardest layer to shortcut — you can't buy a genuine Forbes feature, and you can't manufacture the journalist relationships that produce consistent placements. AuthorityTech's network represents 8–9 years of relationship-building across this landscape.

Layer 2 — Entity Optimization. Structured identity signals that AI engines use to verify who you are: consistent brand entity definitions across authoritative sources, Wikipedia presence, Wikidata entries, Google Knowledge Panel, and LinkedIn company profiles that define the brand in entity-parseable language. Without entity clarity, AI engines can't reliably attribute your earned media to your brand.

Layer 3 — Citation Architecture. Content engineered for AI extraction: proper Schema markup (BlogPosting, FAQPage, Organization), Key Takeaways sections with specific data points, FAQ sections with real follow-on questions, H2 headers that mirror secondary search queries, and inline citations that substantiate every claim. This is the owned content layer — necessary for the foundation, insufficient on its own.

Layer 4 — GEO and AEO. Tactical optimization for generative and answer engines: structured prompts, AI-readable content density, entity-rich intros, comparison tables, and the full suite of generative engine optimization practices. Measuring GEO performance at this layer requires tracking AI Mention Rate, Share of Voice, Citation Frequency, and Prominence Score across multiple AI platforms. Necessary but not sufficient without earned media authority behind it.

Layer 5 — AI Visibility Measurement. Citation frequency tracking across AI platforms: Profound, SiteSignal, and emerging Share of Model (SoM) metrics that track brand authority in AI overviews and chatbot recommendations. This is the measurement layer that has just reached institutional scale. It tells you where you are. The other four layers determine where you go.

PR vs. Machine Relations: What Actually Changed

Dimension	Traditional PR (Pre-AI)	Machine Relations (2024–Present)
Primary target	Human journalists and editors	AI engines + human journalists (both)
Citation goal	Mentions in media that humans read	Placements in publications AI systems cite
Success metric	Impressions, backlinks, brand awareness	AI Citation Frequency, Share of Model, AI Visibility Score
Content type	Press releases, product announcements	Earned placements + citation-engineered owned content
Compounding effect	Short-term (news cycle)	Long-term (training data + RAG retrieval)
Who it convinces	Human readers and editors	AI systems that determine what brands get recommended
The moat	Media relationships (soft moat)	Earned authority footprint (compounding moat)
Payment model	Retainer (pay for effort)	Guaranteed placement model (pay for results)

The shift is not "PR is irrelevant." The shift is: the measure of whether a PR placement works has changed. In the Google era, the primary value of a Forbes mention was the backlink and the branded search lift. In the AI era, the primary value is that Forbes is one of the publications AI training pipelines weight most heavily as a citation authority. The same action — securing a Tier 1 media placement — has dramatically more compound value than it had three years ago. But only if the placement is engineered for citation architecture, includes machine-readable content signals, and is part of a systematic program, not a one-off announcement.

The Case for Moving Now: The Citation Moat Is Forming

Profound's research shows that 34% of AI citations in any given category cluster around a dominant publisher. The authoritative sources AI engines default to are establishing themselves right now — and that selection is driven by which brands have the deepest, most consistent earned media footprint in authoritative publications.

AI search traffic is growing at 9.7x year over year, and the traditional traffic model has fundamentally shifted to a zero-click, AI-first visibility paradigm. The Gartner projection of 25–50% traditional search decline by 2028 is not a long-horizon forecast — it's describing a shift already underway. The brands that are building earned authority now are not being aggressive. They're being timely. The window to establish dominant AI citation share before a category locks is measured in months, not years.

The GEO measurement framework that operationalizes tracking AI citation frequency — across AI Mention Rate, Citation Share, Entity Accuracy Score, and AI Revenue Attribution — gives you the instruments to see the gap. Earned media is what closes it.

Conductor's 2026 research puts this in sharp relief: 32% of digital marketing leaders named GEO their top priority this year, up from 12% in 2025. That's a 167% increase in strategic priority in 12 months. The brands in that leading cohort are not just running GEO campaigns — they're building earned media programs designed specifically to generate the third-party authority signals AI engines weight most heavily.

What This Means for Your Brand Right Now

If you're running a startup that's already gaining traction — and Stripe data shows that more startups are hitting $10M ARR in three months now than ever before, meaning there are more of you than ever — your brand's AI visibility is likely 6–18 months behind your revenue growth. Profound's customers are discovering exactly this: the dashboard shows low citation frequency in categories where the company is already a market leader. The earned media footprint hasn't caught up to the product traction.

The fix is not complicated, but it is systematic. You build earned authority in the publications AI engines weight most heavily. You combine that with citation-optimized owned content that gives the earned media placements somewhere to link to. You track citation frequency, share of model, and AI mention rate. You iterate based on what moves.

AuthorityTech has run this program across 200+ startups and 20+ unicorns, with a 99.9% placement delivery rate across 1,000+ Tier 1 media hits. We operate on a guaranteed placement model: we secure the articles, or you don't pay. The at-risk escrow structure is our answer to the market's fundamental problem — most PR agencies charge $15–25K/month retainers with no delivery guarantee. We charge only when we deliver.

If you want to see where your brand stands in AI search today, the fastest first step is a visibility audit. Run yours here — it maps your current AI citation frequency, identifies the highest-value gaps in your earned media footprint, and gives you a clear starting point for building the earned authority that closes those gaps systematically.

Frequently Asked Questions

Does PR still matter in 2026, or has it been replaced by GEO?

PR matters more than ever in 2026, but the definition of "what makes a placement valuable" has changed. Traditional PR metrics — impressions, backlinks, brand awareness — are now secondary to citation authority: whether an earned media placement appears in publications that AI engines weight heavily when assembling answers. The most effective approach combines GEO-optimized owned content (the foundation) with consistent Tier 1 earned media placements (the accelerant). According to AuthorityTech's data across 200+ client engagements, brands with both layers achieve AI citation rates 10–100x higher than those with owned content optimization alone.

What is Machine Relations, and how does it differ from traditional PR?

Machine Relations (MR) is the discipline of earning AI engine citations and recommendations for a brand — the evolved form of PR for the AI era. Traditional PR convinced human journalists and editors to cover your company. Machine Relations convinces the machines — LLMs, AI search engines, and AI agents — to cite, surface, and recommend your brand when users ask relevant questions. Coined by Jaxon Parrott in 2024, Machine Relations operates across five layers: earned authority (Tier 1 placements), entity optimization, citation architecture, GEO/AEO, and AI visibility measurement. Learn more at machinerelations.ai.

Why do AI engines cite earned media more than brand-owned content?

AI engines are trained to identify and weight authoritative sources — and the defining characteristic of an authoritative source is third-party validation. When an AI system encounters a Forbes article about a company, the signal it receives is: an independent editorial organization with established credibility has determined this company is worth covering. That is a trust signal that brand-owned content structurally cannot replicate. AuthorityTech's analysis shows 82–89% of AI-generated answers cite third-party earned media over brand-owned content. This reflects how AI training pipelines are built, not a temporary market condition.

What does Profound actually tell you, and how do you act on what you find?

Profound tracks how AI models describe and recommend brands across millions of real prompts — showing not just whether your brand appears in AI answers, but why, and what competitors are doing differently. Profound's research shows up to 90% of cited sources in AI answers change over time, meaning AI visibility is an active competitive battleground. What Profound does not do is improve your score for you. The most effective lever for improving AI citation frequency is expanding your earned media footprint in publications that AI engines weight as authoritative — which requires a systematic earned media program, not just content optimization.

How long does it take to build meaningful AI citation authority?

With a systematic earned media program combining Tier 1 placements and GEO-optimized owned content, AuthorityTech clients typically see measurable improvements in AI citation frequency within 60–90 days of the first major placements landing. Brands publishing 12+ citation-optimized pieces per month achieve AI visibility gains 200x faster than those relying on occasional content. Full category authority — the kind that generates 34%-of-category citation dominance — typically takes 6–12 months of consistent program execution. The brands starting now are building moats that will be difficult to close once established.

What's the difference between GEO, AEO, and Machine Relations?

GEO (Generative Engine Optimization) and AEO (Answer Engine Optimization) are tactical optimization practices — structured content, Schema markup, entity signals, FAQ sections — that make brand-owned content more parseable and extractable by AI engines. They are Layers 3–4 of the Machine Relations stack. Machine Relations is the full discipline: it includes GEO/AEO as foundation work, but adds the earned authority layer (Tier 1 earned media placements) that provides the trust signals AI engines weight 10–100x more than brand-owned content. GEO makes you detectable. Machine Relations makes you cited.

Is the guaranteed placement model real, or is it a marketing claim?

AuthorityTech operates on a genuine at-risk model: client funds are held in escrow until placements are secured and delivered. If the placement doesn't land, payment is not released. This model works because AuthorityTech has built journalist and editorial relationships across 8–9 years and 1,000+ Tier 1 placements, achieving a 99.9% delivery rate. It is structurally different from the retainer model most PR agencies use, where payment covers effort regardless of outcome.

Which brands are most at risk of AI visibility gaps right now?

The highest-risk profile is the fast-growing startup with strong product traction but limited earned media history. Stripe's 2026 annual report shows more startups are hitting $10M ARR in three months than ever before — but revenue growth and AI visibility don't compound at the same rate. The dashboard Profound shows these companies typically reveals low citation frequency in their own category, even when they're already a market leader. The second high-risk profile is the established brand that relied on Google search dominance but hasn't built the earned authority signals AI engines prioritize differently than backlinks. Both need a systematic earned media program, not just GEO optimization.