AI Visibility: What Drives It, Where It Comes From, and Where to Invest First
A research-backed guide for marketing, communications, and brand leaders who want to act on AI visibility, not study it.
Marketing, communications, and brand leaders can't escape "AI visibility." It's been everywhere for the past year. Your SEO team is saying it. Conference panels are debating it. And now vendors are selling you dashboards built around it. The conversation has moved fast, gotten noisy, and produced almost nothing useful for making an actual decision.
This paper is here to change that.
It's written for people who want to act on AI visibility, not study it. By the time you finish reading this, you'll be able to do four things:
- Explain what AI visibility is in plain English.
- Identify the three categories of signals that determine whether AI tools mention your brand.
- Learn which category matters most and why.
- Decide where your company should invest next.
What AI visibility is and why it matters
Search has changed underneath us. The competitive question is no longer where you rank, but whether your brand surfaces inside the answer.
What AI visibility is
For twenty years, search followed a predictable pattern. A user typed a question, Google returned ten blue links, the user clicked one, and if your page appeared on that list, you had a chance at the conversion. The goal was a ranking, and success was a click.
AI-mediated search breaks that pattern. A user still types a question, but what comes back is an answer instead of a list. Sometimes it arrives as a direct recommendation ("the best CRMs for mid-market teams are X, Y, and Z"), or as a summary with citations below. It can also play out across three or four exchanges before the user lands on anything. The user might click through to a website at the end of that conversation, but often they won't, because the decision has already been made.
Ownership has changed with it. Where pages were squarely a marketing concern, sources can come from anywhere. A journalist quoting your CEO, a Reddit thread recommending your product, a customer review, an analyst report, a podcast transcript, your own site: all of these now feed the same machine, and the machine doesn't care which department produced them.
Throughout this paper, "AI visibility" refers to two related things. Both matter, and most organizations underinvest in both.
Why AI visibility matters commercially now
Three trends make this urgent. Each is backed by research. Each is already commercial.
Consumer adoption has reached the point where AI search is a revenue channel, not a novelty. Google's AI Overviews, the summary boxes that appear above the blue links, reached 2 billion monthly users across more than 200 countries and 40 languages by mid-2025. McKinsey projects that $750 billion in U.S. revenue will flow through AI-powered search by 2028, which is roughly three-quarters of the total search market.
B2B buying has moved faster than most executives have noticed. G2 research published in early 2026 found that 51% of B2B software buyers now open an AI chatbot before they open Google when they begin vendor research. A year earlier, that number was 29%. Research, shortlisting, and vendor comparison are now happening inside AI tools, with or without your knowledge.
Traditional search is losing ground. Ahrefs, one of the largest SEO and marketing platforms tracking hundreds of thousands of keywords over two years, found that when an AI Overview appears above a query, the top-ranked organic link loses up to 58% of its click-through rate compared to the same position without one. Year after year, as your team invests in ranking for commercial queries, the value of that ranking quietly erodes.
Taken together, the picture is straightforward. Buyers are shortlisting inside AI. Your brand may or may not make the cut. And the blue links that used to catch the rest are sending less traffic every month. The companies that'll benefit most are the ones that recognize the shift is already here.
What drives AI visibility
Three signal categories shape every AI recommendation. They aren't equal in weight. Investing in them in the wrong order is the most common and most expensive mistake we see.
The three categories of AI visibility signals
When AI systems form a recommendation, they draw on three distinct pools of evidence. Each pool is populated by a different kind of work, owned by different teams or vendors, and weighted differently by the algorithm. Investing in them in the wrong order is the most common and most expensive mistake we see. The instinct to invest in all three is a sound one.
These categories aren't equal. Research spanning millions of AI citations consistently shows the same ordering: earned authority dominates, consensus signals come second, and owned clarity sits beneath both.
Why earned authority carries the most weight
What makes this notable is how consistent the evidence is, especially for a field this young, and especially given that the researchers involved have no incentive to reach the same conclusion.
Muck Rack analyzed more than a million links cited by ChatGPT, Gemini, and Claude across hundreds of thousands of queries. They found that 82% of AI citations came from earned media, and 94% came from non-paid sources altogether. Omniscient Digital ran a separate analysis across five AI platforms, examining 23,387 unique citation sources, and reached a compatible conclusion: for queries where users mentioned a brand by name, earned media accounted for 48% of citations while owned brand content supplied only 23%. For customer-review queries, the earned share jumped to 82%.
AirOps looked at the same question through the lens of brand-mention volume, analyzing 21,311 brand mentions across ChatGPT, Claude, and Perplexity. Only 13% of mentions came from the brand's own domain. The remaining 85% lived on external sites, and nearly nine in ten of those external mentions appeared inside listicles, comparison posts, or reviews written by someone other than the brand itself.
Ahrefs ran the study that makes the case most directly. Looking at 75,000 brands, they measured how different factors correlated with whether an AI system cited the brand in its answers. Mentions of a brand's name on the open web showed three times stronger correlation with AI visibility than traditional backlinks. The top three correlating factors were all off-site brand signals. Backlink metrics, which dominated SEO for two decades, barely registered.
Why the pattern holds so consistently
-
Trust inheritanceAI systems borrow credibility from the publications they cite. A mention of your CEO in the Wall Street Journal does more than reach the Journal's readers. It signals to the model that a trusted source took your organization seriously. This raises its confidence in citing you on related topics down the line.
-
Self-promotion detectionAI systems are trained to recognize and discount marketing copy. When a page sounds like a brochure, the model deprioritizes it, regardless of factual accuracy. Editorial voice is baked into the way large language models assess source quality.
-
Cross-source corroborationA single claim on your site is weaker evidence than the same claim echoed by a journalist, an analyst, and a customer review. This kind of distributed reinforcement is something earned media produces naturally and something owned content, by its very nature, will never replicate.
What this means in practice is that most organizations have their AI visibility investment backwards. They're optimizing what's inside. The thing that matters most is outside. And most organizations aren't touching it. Correcting that imbalance is the single highest-leverage move available to most brands today.
Why reputation and consensus signals are the second-heaviest input
Earned authority is what's on the record. Consensus signals are what people say off it. AI systems draw on both, and the distinction matters because the two require very different kinds of work.
Semrush's three-month analysis of 230,000 prompts and more than 100 million AI citations across ChatGPT, Google AI Mode, and Perplexity found that the web's consensus layer sits at the center of AI recommendation. Reddit remained the top-cited domain on Perplexity and held a top-three position on both ChatGPT and Google AI Mode, with Wikipedia and YouTube anchoring the rest of the leaderboard. Even as platform updates reshuffle individual citation shares, the underlying pattern holds. ChatGPT's September 2025 sourcing rebalance sent Reddit's citation rate from nearly 60% down to around 10%. Wikipedia's fell from roughly 55% to under 20%. Reddit still held a top-two spot overall.
That finding is reinforced by commercial behavior on the platform side. Google paid Reddit $60 million per year for content access, and OpenAI signed a similar licensing deal with Reddit valued at roughly $70 million annually. That's not spin, that's how the system is built. The AI companies bought the conversation because the conversation is what their models need.
Reviews sit inside this same category and carry comparable weight. BrightLocal's consumer survey found that 97% of consumers still read reviews before making a decision, and roughly nine in ten said reviews meaningfully shape their overall perception of a chain or franchise. AI reads your reviews. When someone asks ChatGPT about your category, your reputation in the model is partly just your last two hundred reviews, compressed.
Multi-location and local brands face the sharpest version of this problem. SOCi analyzed 350,000 business locations and found that only 1.2% were recommended by ChatGPT, compared to 35.9% appearing in Google's local 3-pack, the boxed map listing that highlights three nearby businesses on a local search. AI local visibility runs roughly thirty times more selective than traditional local search. Winning the 3-pack no longer guarantees a place in the AI recommendation, and the gap between the two is widening.
One reason consensus signals carry the weight they do is that they're genuinely difficult to fake. AI discounts fake reviews, paid forum posts, and engineered narratives. It's gotten good at spotting them. The only reliable way to build consensus authority is the slow one: earning it across dozens or hundreds of independent voices over time.
This is the work most organizations are structurally worst at. Review platforms aren't typically PR's job, community engagement isn't SEO's, and narrative work doesn't usually sit with CX. Your departments have gaps. Your reputation lives in them. AI doesn't see the gaps, it just sees the signal.
Why owned content is foundation, not differentiator
A useful way to understand this category is through the order of operations. Earned authority and consensus signals determine whether an AI system considers you for a recommendation. Once the model decides to cite you, owned content determines which pages it reaches for.
The question of what makes a page citation-worthy has been studied with some rigor. The most-cited original research in the field is the GEO paper from Princeton, Georgia Tech, and IIT Delhi, which tested structured optimization techniques across ten thousand queries. The findings were specific and actionable:
Notice that the single highest-lift tactic, adding third-party quotations, is a downstream version of earned-authority work. The pages that win most in AI citation are the ones built around credible outside voices, not the ones that talk about themselves more.
Kevin Indig's analysis of three million ChatGPT responses and thirty million citations surfaced a structural finding worth acting on. 44% of ChatGPT citations come from the first 30% of a page. If the answer to the user's question is buried below the fold, the AI rarely reaches it. Pages that front-load their substance get cited, and pages that ramp slowly toward a payoff get skipped.
Get the technical layer wrong and nothing else matters. Get it right and it stops being a factor. Schema markup, clean HTML, server-side rendering, AI bot crawlability, fast load times, entity consistency across properties, none of it will make an AI system recommend your brand on its own. All of them, absent or broken, can disqualify you from being read at all.
There's one more pattern worth naming. Content volume doesn't correlate with AI visibility. Publishing twenty blog posts a month moves nothing if the pages are shallow, promotional, or poorly structured. A single authoritative page, dense with statistics and expert commentary, outperforms a library of generic content. The content-factory approach that drove a decade of SEO output is now actively counterproductive for AI citation.
For most organizations, owned content already consumes the majority of marketing investment. The research makes clear that the ratio is wrong. Clean up owned content. Then stop over-investing in it. The signal weight lives in earned and consensus work.
What to do about it
A map of which agency type influences which signal, why PR and reputation firms hold structural advantage, and a 90-day plan for organizations that decide to act.
Which agency types influence which parts of AI visibility
AI visibility doesn't fit neatly inside any one budget, leader, or agency. No vendor owns all of it. Know which firm influences which layer or risk asking the wrong agency to solve the wrong problem.
| Agency type | Strongest in | Typically weaker in |
|---|---|---|
| SEO and content agencies | Owned clarity, topical coverage, and on-page structure | Earned authority, executive visibility, reputation repair, external validation |
| Technical SEO & CRO specialists | Accessibility, site performance, conversion architecture | Narrative, editorial, reputation |
| PR and communications firms | Earned authority, executive positioning, narrative, crisis framing | Technical SEO, on-site implementation, review ops |
| Reputation management firms | Managing reviews, tracking sentiment, repairing public perception | Technical SEO, editorial depth (varies by firm) |
| CX and review platforms | Review generation and response workflows, feedback loops | Broader authority work, media relations |
| Local SEO and multi-location vendors | Directory listings, Google Business Profile, maps, location pages | National authority, executive narrative, PR |
| Full-service comms firms with AI fluency | Coordination across the highest-leverage layers | Not every tactic needs to be in-house |
The gap most companies fall into looks something like this. One or two vendors get hired. The rest gets ignored. Months later, the investment is stacked in one or two categories and the third is untouched. The fix isn't more spending, it's better distribution.
Where PR and reputation firms hold a structural advantage
The framework makes one thing clear. The two signals that matter most to AI systems, earned authority and reputation consensus, are also the two that organizations are least equipped to produce on their own.
You can write your own content, but you can't credibly quote yourself in the Wall Street Journal. You can build your own site, but you can't assign your customers a script for their reviews. You can publish a thought leadership post under your CEO's byline, but you can't manufacture the analyst report that references her or him three years later. The inputs AI systems trust most are, by design, the ones your internal team can't manufacture. That's not a coincidence, it's precisely why those inputs carry the weight they do.
PR and communications firms are built for this. Their core work produces exactly what AI systems are looking for. Relationships with journalists, executive positioning, message discipline across external touchpoints, crisis framing, and reputation repair all generate material that AI systems read as third-party validation. What PR and communications firms actually do hasn't changed. What's changed is how the results get measured. The placement in a trade journal that used to drive a monthly press-clip report now also feeds an AI model's view of what your brand is authoritative in, and continues doing so for as long as that article lives on the indexed web.
Reputation management firms are just as well positioned. Reviews, sentiment, and crisis coverage all feed the consensus layer. AI reads that layer the same way it reads editorial. The strongest reputation firms operate across both the formal record and the informal one, which gives them a broader lever than firms working only one side.
None of this means a PR or reputation firm should execute every tactic in the stack. Technical SEO, web development, conversion optimization, and platform-specific review management are all better served by specialists. It's not about how much you do. It's about what actually moves the needle. Specialists handle the supporting inputs. The comms and reputation firm handles the heaviest ones, and coordinates the rest so the system compounds in one direction.
The AI visibility maturity model
Not every organization needs to start in the same place. The five-level model below describes the trajectory from invisible to category authority, and the right next investment depends on which level you're at now.
Most mid-market organizations sit between Level 2 and Level 3. Most enterprises sit between Level 3 and Level 4. Level 5 is rare, and once achieved, very hard for competitors to dislodge.
Where to invest first
Skip the checklist. The right move depends on your specific problem. Four patterns cover almost every situation we see.
- Likely root causes
- Weak earned-media footprint, thin topical coverage on your own site, limited third-party validation of your expertise, inconsistent entity signals across the web.
- Where to invest first
- Earned authority takes priority. A targeted PR and thought-leadership program aimed at the specific publications and analysts your category cites most. On-site content cleanup comes second. Technical SEO comes third, unless something is actively broken.
- Likely root causes
- Stale or fragmented brand narrative, crisis residue still living in the training data, inconsistent messaging across site, press, and executive commentary, negative coverage that outweighs recent positive coverage.
- Where to invest first
- Reputation and narrative work. Coordinated messaging refresh across owned properties, earned media, and executive channels. Suppression of outdated content where appropriate. A steady cadence of fresh, authoritative commentary that replaces the dominant narrative over time.
- Likely root causes
- Review quality, velocity, or recency below the AI threshold. Inconsistent profile data across platforms. Weak local content beyond the basic Business Profile. Limited location-level authority signals.
- Where to invest first
- Review ecosystem work, starting with generation velocity and response workflows. A profile data audit across the full directory ecosystem, not just Google. Location-level content on your owned site. A review platform partner if one isn't already in place.
- Likely root causes
- Trust fragility, regulatory and media exposure, stakeholder complexity, long-tail crisis memory that persists across model versions and training refreshes.
- Where to invest first
- A strategic communications partnership with deep industry experience. Executive media training and positioning. Proactive reputation monitoring with defined escalation paths. A sustained narrative infrastructure of spokespeople, thought leadership, and crisis-ready messaging that compounds over time.
The first 90 days
For organizations deciding to act, the first ninety days should establish a baseline, align internal ownership, and put the first authority-building work into motion. A practical starting sequence looks like this.
- Run a branded prompt audit. Ask ChatGPT, Gemini, Perplexity, and Claude the questions your buyers ask. Document what the tools say about you, what's missing, and how you compare to two or three competitors.
- Run a non-branded prompt audit. Ask the same tools your category's shortlist questions ("best X for Y"). Document whether you appear, where, and in what company.
- Assign internal ownership. AI visibility cuts across PR, SEO, content, and CX. One senior leader needs to own the overall picture, even if execution stays distributed.
- Address technical disqualifiers on the owned-content side. Crawlability for AI bots, schema on core pages, entity consistency across your site, social profiles, and major directories.
- Audit your earned-media footprint against where AI tools actually pull citations from in your category. Most organizations discover a meaningful gap at this stage.
- Audit your review profile on the platforms that matter for your industry. Volume, velocity, recency, sentiment.
- Launch a targeted thought-leadership program aimed at the specific publications your category's AI answers cite most. Quality over volume.
- Begin a review ecosystem program if the audit surfaced gaps worth closing.
- Put monitoring in place so you can see what's improving, what isn't, and where the model's view of your brand is shifting month over month.
The goal at the end of ninety days isn't to have won AI visibility. The goal is to have a clear picture of where you stand, internal ownership of the problem, a foundation that's no longer actively holding you back, and the first wave of authority work in market.
AI visibility is the compounded outcome of authority, reputation, clarity, and distributed trust. The organizations that'll win in AI-mediated discovery are the ones that treat it as a cross-functional brand-building imperative rather than a technical SEO problem.
The weighting of the inputs is clear enough in the research to act on. Earned authority, the third-party coverage and expert commentary that you can't manufacture from inside your own walls, carries the most weight. Reputation and consensus signals come next, and their influence is nearly as heavy. Owned clarity is necessary but insufficient on its own. That ordering isn't an argument against any one kind of work. It's an argument for getting the order right, because most organizations today invest heavily in the signal category with the least weight, moderately in the middle, and almost nothing in the heaviest.
The window for establishing advantage in this category is narrow. Every month the training data accumulates, brands cited often get cited more, and companies invisible today become harder to reach tomorrow. The accumulation effect is why early action matters more than scale of action.
Solv Communications helps organizations operate at the authority, reputation, and narrative layer of this system. We don't replace your SEO team, your web developer, or your review platform. We handle the highest-leverage inputs directly and coordinate across the rest, so the full system compounds in a single direction.
Reference material
A.Glossary of terms
- AI visibility
- How often a brand is mentioned, cited, or recommended when a user asks an AI tool a question. Umbrella term covering both branded and non-branded AI queries.
- AI brand visibility
- What AI tools say about a specific company when asked about it by name. Shaped by consistent messaging across owned, earned, and reputation sources.
- Non-branded AI visibility
- Whether a brand appears in AI answers to category-level queries ("best X for Y"). Where shortlists are formed.
- AI Overview (AIO)
- The AI-generated summary Google places above the blue links on a search results page. Powered by Gemini.
- AI Mode
- Google's separate conversational AI search interface, distinct from AI Overviews. Uses "query fan-out" to research sub-questions in parallel.
- Citation
- An AI tool's reference to a specific source in its generated answer. Replaces ranking as the primary unit of AI search visibility.
- Earned media / earned authority
- Third-party coverage, quotes, or mentions of a brand that the brand did not directly pay for or publish. Press coverage, analyst reports, editorial features, podcast appearances.
- Owned content / owned clarity
- Content a brand publishes on properties it controls. Website pages, blog posts, case studies, product pages.
- Reputation and consensus signals
- Distributed public sentiment across reviews, forums, community platforms, and ratings. What the web collectively says about a brand's trustworthiness.
- Entity
- A distinct concept, person, company, or thing recognized by AI systems as a unique identity. Brands work to be recognized as consistent entities across the web.
- GEO
- Generative Engine Optimization. The practice of optimizing content for citation in AI-generated responses. The AI-era counterpart to SEO.
- LLM
- Large Language Model. The foundational technology behind AI search tools. Models trained on large volumes of text that generate responses by predicting likely next tokens.
- RAG
- Retrieval-Augmented Generation. An AI architecture that retrieves fresh web content in real time and uses it to generate answers. Perplexity is fully RAG-based; ChatGPT uses RAG for some queries.
- Schema markup
- Structured data added to web pages in a format AI systems can parse cleanly. Common types include Article, FAQPage, Organization, Product.
- llms.txt
- A proposed standard for a plain-text file placed at a domain root that provides AI-friendly content guidance. Analogous to robots.txt for search crawlers. Adoption is growing, though no major AI platform has formally committed to reading it.
- E-E-A-T
- Google's quality framework: Experience, Expertise, Authoritativeness, Trustworthiness. Originally a search ranking concept, now also used to describe signals AI systems evaluate.
- Training data
- The corpus of text AI models learn from during training. Content in training data shapes what models "know" about brands before any live search happens.
- Zero-click search
- A search that ends without the user clicking any result. Rising sharply with the spread of AI Overviews and answer boxes.
- Agentic commerce
- Transactions initiated or completed by autonomous AI agents on behalf of a buyer, rather than manually by a human.
- Prompt audit
- Structured testing of what AI tools say about a brand or category. Usually involves running a defined set of branded and non-branded prompts across multiple AI platforms and documenting the outputs.
- Self-promotion detection
- The pattern by which AI systems recognize and discount marketing or brochure-style content in favor of editorial voice.
B.Extended statistics
These statistics didn't make the main narrative but are accurate, well-sourced, and useful context if you want to go deeper on any section.
On market context and adoption
On earned authority
On reputation and consensus signals
On owned clarity and technical signals
C.Full source list
Sources cited in the main paper, in order of appearance.
- TechCrunch — Google AI Overviews reach 2B monthly users
- McKinsey — New Front Door to the Internet
- G2 / PR Newswire — 51% of B2B buyers start research with AI chatbots
- Ahrefs — AI Overviews Reduce Clicks (updated study)
- Muck Rack — Generative AI Relies Heavily on Earned Media and Journalism
- Omniscient Digital — How LLMs Source Brand Information
- AirOps — The Influence of Offsite Signals in AI Search
- Ahrefs — Top Brand Visibility Factors (75,000 brands studied)
- Semrush — The Most-Cited Domains in AI (3-Month Study)
- BrightLocal — Local Consumer Review Survey
- Search Engine Land / SOCi — AI Local Visibility Report 2026
- Aggarwal et al. — GEO: Generative Engine Optimization (Princeton, Georgia Tech, IIT Delhi)
- Search Engine Land / Kevin Indig — 44% of ChatGPT citations from first third
- Solv Communications — Contact
