The question now isn’t how to rank in Google, but how to be quoted by the machines – and this piece explores what drives citations, which formats get lifted, and how marketing and PR can measure success in a world without clicks.

For two decades, marketing and PR professionals have obsessed over rankings and clicks: secure a high Google position, win traffic, measure ROI. That formula has now fractured. Generative AI increasingly gives users complete answers without a click: in Google’s AI Overviews, in ChatGPT’s browsing mode, or in Perplexity’s extended recommendations. Brands are either named in those answers or erased from the user’s frame of consideration.

The numbers show the scale of disruption. Research from BrightEdge indicates that click-through rates for top organic results can fall from around 25% to as low as 7% when an AI Overview appears. At the same time, Pew Research reports that one in five US adults now uses AI assistants for search weekly, with Gen Z already above 30%. For international projects and outsourcing categories, the effect is especially stark: many prospects may never see a traditional Google results page at all. They ask ChatGPT or Perplexity: “Who are the best agencies in Lithuania?” and take the machine’s answer as reality. If your brand is missing, you’ve lost before the click.

The recipe from the winners

Salesforce provides a textbook example. Ask ChatGPT or Claude to name the leading CRM platforms, and Salesforce invariably appears at the top. This prominence is not accidental. For years, the company has invested in a meticulous Wikipedia presence, analyst reports, case studies, and constant coverage in business media. It is precisely the kind of brand that AI models recognise, because its digital footprint is wide, trusted, and structured. Salesforce “seeded” the ecosystem before LLM-seeding had a name.

HOKA, the running shoe brand, illustrates how even challengers can break through. Once a niche player, it systematically cultivated reviews, athlete endorsements, and coverage across sports media and YouTube. Now, when users ask Perplexity or Google AI for “the best running shoes for long distance,” HOKA regularly appears alongside Nike and Adidas. Community chatter, long-form reviews, and expert lists transformed it into an AI-visible leader.

These cases prove a critical point: generative search does not only reward scale. It rewards presence across the trusted sources that models draw from. Marketers have woken up to the same realisation: traditional SEO is no longer sufficient. Ranking on page one does not guarantee you’ll be cited in an AI answer. This is why agencies are rushing into audits, companies are refreshing their PR pipelines, and conversations about “AI visibility” are appearing on every marketing agenda.

The argument that brand awareness is “immeasurable” and therefore less important than performance marketing no longer holds. Generative search has made awareness measurable again – in the form of citations. If 30% of AI answers in your category include your competitor and not you, the performance gap is real, quantifiable, and urgent.

How do different AI systems choose their sources?

Although each platform has its own mechanics, together they reveal how authority and structure shape visibility.

ChatGPT, when connected to the web, behaves less like a broad conversationalist and more like a selective curator. According to Profound’s analysis of nearly 30 million citations from August 2024 to June 2025, ChatGPT’s top-10 most cited sources are overwhelmingly dominated by Wikipedia (47.9%), with Reddit trailing at 11.3%. Following those, sources like Forbes, G2, TechRadar, NerdWallet, BusinessInsider and Reuters appear – while brand-owned blogs and product pages are barely present. In other words, ChatGPT elevates encyclopaedic and journalistic authority above community voices and promotional content. If your brand wants to be quoted, being present in independent, high-trust sources is no longer optional.

Google’s AI Overviews, by contrast, are deliberately eclectic. BrightEdge research shows YouTube as the single most cited domain, followed by blogs (39%) and news publications (26%). Reddit, LinkedIn, and Medium also surface regularly. Here, the principle is not conservatism but balance: SGE blends expert articles with authentic community input. Interestingly, Google often privileges fresh and structured content – updated guides, clear lists, FAQ sections. Compared to ChatGPT’s deference to traditional media, Google’s model is more willing to treat a Reddit thread or a LinkedIn post as a meaningful source, provided it feels credible and recent.

Perplexity takes eclecticism to an extreme. Profound’s leaderboard data shows that nearly half (46.7%) of its top citations come from Reddit, with another 14% from YouTube, alongside frequent mentions of review platforms like G2, PCMag, TripAdvisor, and Yelp. Unlike ChatGPT, Perplexity thrives on community conversation and detailed user-generated reviews. It also has a stylistic quirk: it loves long lists. While ChatGPT or Google AI may cite 3-8 brands, Perplexity often produces 10-15 recommendations. For marketers, that makes affiliate-style content – top-10 comparisons, ranked reviews, and best-of lists – especially powerful. The model is predisposed to surface those formats, giving challenger brands a way to appear alongside incumbents if they are visible in the forums, reviews, and rankings Perplexity trusts.

Claude, developed by Anthropic, sits closer to GPT. It tends to prefer encyclopaedias, high-quality news, and factual reports, producing calm, neutral summaries. Compared to Perplexity’s embrace of Reddit and YouTube, Claude is more restrained. The implication for brands is clear: visibility here relies less on community buzz and more on appearing in professional, well-documented contexts.

What kinds of content get quoted?

Across all four platforms, patterns emerge. Generative models are not “reading” like humans; they are scanning for extractable facts. The formats that surface most often are those that make this extraction easy.

Structured lists and rankings are prime examples. Articles like “Top 10 CRMs for 2025” or “Best running shoes for marathons” provide exactly the kind of concise, comparable snippets that models reuse. Being included in such lists, especially on authoritative domains, significantly increases the chance of citation.

Q&A-style content also maps neatly onto user queries. FAQs, interviews, and comparison pieces framed as “Which is better, A or B?” are lifted readily. When the format mirrors the natural phrasing of a question, the model sees it as a ready-made answer.

Freshness is another consistent factor. Google’s AI Overviews in particular show a bias toward recently updated material. Articles marked “Updated in 2025” or guides refreshed with new data are far more likely to appear than static, older pages. Likewise, recent news stories on authoritative outlets are incorporated when queries touch current events.

Finally, structural clarity matters. Content with headings, sub-sections, and tables is easier for models to parse and quote. The HOTH’s research by Rachel Hernandez confirms that clean HTML structure and subheadings measurably improve citation likelihood. For brands, the lesson is straightforward: write with machines in mind, as much as people.

Tactical steps for brands

The first step is opening the gates. Too many brands still block AI crawlers in robots.txt, whether by oversight or conservatism. OpenAI’s GPTBot, Anthropic’s ClaudeBot, and PerplexityBot all need to index your site. If they are excluded, you have already removed yourself from contention.

Next comes auditing your current presence. This is best done through “blind audits”: ask each AI platform the questions you would want to be cited for, and record the answers. Who is named instead of you? How are you described? The tone and context matter as much as the mention itself. Manual audits can be done weekly, while platforms like BrightEdge AIO, Semrush AIO, Profound, and Promptmonitor offer automated tracking of “share of AI voice.”

Publishing where AI looks is the strategic heart of LLM-seeding. Your own website is not enough; models draw heavily from high-authority ecosystems. That means placing content on Medium, LinkedIn, and Substack, contributing to niche industry media, securing listings on G2 or Capterra, cultivating reviews on TripAdvisor or Yelp, and authentically participating in Reddit and Quora discussions. YouTube deserves special emphasis: Google’s SGE cites it heavily, and a single well-optimised video can appear directly in an AI Overview. Wikipedia, though harder to influence, remains disproportionately powerful. A verified entry is a gateway to recurring citations across GPT, Claude, and Google AI.

Optimising for machine readability is the next discipline. Structure your content with clear headings, straightforward language, and extractable sentences. Build FAQs and Q&A sections directly into your sites and blogs. Consider tables and comparative matrices – not for aesthetic design, but because they make facts more digestible for machines.

Finally, do not abandon the basics. Around half of all AI citations still come from sites that already rank in Google’s top 10. Classic SEO, authoritative backlinks, and earned media coverage continue to feed AI visibility. Digital PR – placements in respected outlets, industry awards, analyst reports – now pays double dividends: it persuades humans and teaches machines.

New metrics for a citation economy

The way success is measured must also evolve.

Share of AI Voice is emerging as a core KPI, analogous to traditional share of voice in media monitoring. If your competitor appears in 30% of AI answers in your category and you appear in 10%, the gap is immediate and quantifiable.

AI Visibility Score, weights frequency, position, and prominence of mentions. Being the first brand cited in a Perplexity list carries far more impact than being the twelfth.

Quality and sentiment of sources also matter. A citation from Wikipedia or the Financial Times has more weight than one from a low-quality blog, and how the brand is described in that source will shape the AI’s output. PR teams are beginning to treat AI mentions with the same sentiment analysis once reserved for press coverage. In practice, this might mean setting a quarterly goal such as “+15 new mentions in top-ranked industry outlets or review platforms."

AI referral traffic is another frontier. While links inside AI answers are less frequently clicked, analytics can detect referral spikes from AI user agents or from correlated increases in branded search. These indirect signals help connect AI visibility to tangible outcomes.

Finally, AI mentions tracking is becoming a discipline of its own. Agencies are including “AI clippings” in client reports: how many times a brand was named by ChatGPT, Google AI, or Perplexity over a given period, in what context, and alongside which competitors.

PR and AI want the same thing

For years, PR struggled to prove its worth against the metrics of performance marketing. Generative AI has rewritten the rules. Visibility is no longer only about driving traffic, but about being named inside the answer itself.

The surprising truth is that PR and AI share a goal: to deliver credible, structured, trustworthy information. Brands that adapt early – by opening access, seeding trusted domains, structuring content for machine readability, and measuring AI-specific visibility – will quietly accumulate presence across ChatGPT, Google AI, Perplexity, and Claude.

As AI search usage grows from today’s 20% toward 30-40% in the coming years, those brands will become the “default answers.” Competitors will find themselves trying to dislodge names that the machines already trust.

In the age of generative search, the currency of PR is no longer clicks. It is citations.