The Top 20 Sources AI Cites When Candidates Research Employers

When a candidate asks ChatGPT, Perplexity, or Google AI "what's it like to work at this company?" — the AI doesn't make up an answer. It cites.

We analyzed 100,000+ AI responses to employer-brand-related prompts, captured across the AI engines that disclose their sources at response time: Perplexity, Google AI Overviews, OpenAI Search, and Google AI Mode. 78,311 of those responses included at least one citation. Every cited URL was captured, every domain mapped to a canonical platform, and every response sliced by the type of question being asked.

This is the resulting picture: the actual sources shaping AI's narrative about employers, and what employer brand teams need to understand about a citation graph they don't yet control.

Candidates have already moved. According to ZipRecruiter's New Hires Survey, more than half of recent hires used generative AI during their job search — a figure that doubled in a single year. Gartner predicts traditional search volume will drop 25% by 2026 as AI takes over the answer layer. Employer brand teams are now competing for citation share on a surface they don't control — and most don't yet know what they're being cited as.

The 20 most-cited AI employer brand sources

Three patterns inside the ranking

Glassdoor isn't dominant. It's hegemonic.

Present in more than one in three cited AI responses about a major employer — more than the next two platforms combined. Across every employer in the dataset, Glassdoor's presence rate never drops below 20%. No other source comes close to that floor. If a candidate is asking AI about an employer and seeing a citation, statistically, that citation is most likely Glassdoor.

The community tier rivals every dedicated employer brand platform combined except Glassdoor.

Reddit, YouTube, Facebook, Instagram, and Quora together account for 30% percentage points of presence — nearly Glassdoor-sized in aggregate. None of it is on any employer brand optimization roadmap. None of it is "managed" the way a careers site is managed. AI is reaching for user-generated, lived-experience content at near-Glassdoor scale, and most employer brand teams are still optimizing for the things they can control.

Below the universal players, the long tail is regional and specialty-driven.

AmbitionBox dominates India. Kununu owns the German-speaking markets. Levels.fyi specializes in compensation transparency. IGotAnOffer, Exponent, and Taro live almost entirely in tech interview prep. Forbes, Fortune, and Great Place to Work surface for rankings-driven discovery. The "global" view obscures these specialists — but for any individual employer, the right mix depends on industry, geography, and the question being asked.

Different question, different sources

Candidates don't ask AI one question. They ask four — and the source mix shifts dramatically depending on which.

The four-lens framework breaks candidate intent into four distinct modes:

• Discovery — "What companies should I be looking at as a [role] in [location]?"
• Experiential — "What's it actually like to work at [company]?"
• Competitive — "How does [company A] compare to [company B]?"
• Informational — "What does [company] pay? What are the benefits?"

Discovery: where rankings and tech career hubs earn their keep

When AI is generating recommendations — what companies a candidate should consider — it pulls heavily from curated rankings and tech career hubs. Great Place to Work jumps to second place, more than doubling its global presence. When the question is "companies you should consider," AI reaches for editorial rankings rather than employer review platforms. Built In hits its peak rank here too, well above its global standing — Discovery is its natural habitat. Indeed, the global #2, drops all the way to sixth — because job-board search isn't the same as candidate discovery. Indeed is where you go when you know what you want. Discovery citations come from sources that surface options.

Experiential: when culture is the question, employee voice wins

Asked what it's actually like to work somewhere, AI concentrates around dedicated employer review platforms and lived-experience content. Indeed jumps to a strong second on the back of its review section, not its job postings. Comparably surfaces in the top three. Reddit and YouTube both appear in force as candidate-told stories about day-to-day work. AmbitionBox and Kununu carry weight regionally. The pattern is consistent: when AI is asked to characterize culture, it reaches for platforms specifically built for employee voice. Tech career hubs and ranking sites — including Built In and Great Place to Work — mostly disappear here.

Competitive: head-to-head questions live in community channels

When the question is "X versus Y as employers," AI flips its source mix toward unstructured community content. Reddit, YouTube, Quora, Facebook, and Instagram collectively account for 43% of all competitive citations — bigger than any single dedicated employer brand platform, and bigger than the entire dedicated-platform tier combined. Comparison questions don't get answered with curated content. They get answered with candidate voice in candidate channels. This is the lens where the limits of "official" employer brand content show up most clearly.

Informational: where compensation tools finally surface

Pay, benefits, interview process, day-to-day facts — this is where compensation tools surface meaningfully. Levels.fyi nearly doubles its global average here — its highest performance across any lens. IGotAnOffer emerges for interview-process questions. The shift is real: when AI is answering structured factual questions, it reaches for sources built around structured factual data. But even at its peak, Levels.fyi only ranks seventh. The primary salary signal across the dataset is still review platforms with self-reported pay data, not specialist comp tools.

The implication is that each lens has a natural cast of characters. Built In and Great Place to Work matter for Discovery and disappear from Experiential. Levels.fyi matters for Informational and almost nowhere else. Reddit and YouTube own Competitive. Optimizing for any single platform without specifying which lens you're targeting is buying coverage of roughly one quarter of the AI citation graph.

Regional fragmentation

The "global" view obscures regional reality. Outside the universal top tier, AI's citation behavior is geography-dependent — sometimes dramatically.

India: AmbitionBox is the de facto second platform

For any employer with meaningful India presence, AmbitionBox is the default second platform after Glassdoor. It appears in citations for nearly every employer in the dataset and reaches its highest presence for India-focused queries. For India-specific role questions, AmbitionBox's footprint can outweigh Comparably, Built In, and Levels.fyi combined.

DACH: Kununu owns the German-speaking markets

For German-headquartered or Germany-recruiting employers, Kununu's presence climbs into double digits and far higher for German-language queries. Outside DACH, it barely appears. A DACH employer brand strategy that ignores Kununu is the regional equivalent of ignoring Glassdoor.

Japan and Korea: the citation graph reorganizes entirely

For Japan- and Korea-specific queries, the citation graph reorganizes around regional platforms — Note.com and OpenWork in Japan, Naver Blog, Brunch, JobPlanet, and Saramin in Korea. For employers recruiting in these markets, these platforms can outweigh the entire global top 10. Excluding them from a regional employer brand strategy is the equivalent of excluding Glassdoor from a US strategy.

LATAM, MENA, and emerging markets: the long tail thins

In LATAM, MENA, and other emerging markets, the citation graph thins out. Glassdoor and Indeed cover most of LATAM. The MENA region depends heavily on LinkedIn and local job boards. AI's citation graph in these regions is structurally less developed, which is itself useful context — there's whitespace for employers to influence narrative before that fragmentation hardens.

The regional pattern reinforces the conclusion from the lens and industry views. A single-platform AI employer brand strategy is, structurally, a single-geography strategy. The platform that works in San Francisco doesn't move the AI conversation in Bangalore or Berlin or Tokyo.

What this means for employer brand strategy

The strategic question isn't "which AI employer brand platform should we optimize for?" It's "what's the citation profile we need across all four lenses, in our actual markets, for our actual functions — and where does our current presence have gaps?"

The right mix depends on the candidate, the question, the function, and the geography. There is no single answer. There is no platform that handles all four lenses, all major industries, and all major regions at scale. The citation graph specializes in every direction the data is sliced.

Any pitch that begins "we're the platform for AI employer brand optimization" needs to answer one question: which lens, which industry, which region? Most platforms are dominant in zero. The handful that lead a single lens lead by single-digit margins over the next contender. Optimizing for any single source — even one as universal as Glassdoor — covers a slice of the citation surface area, not the whole.

The actual strategic problem is measurement and routing. Knowing which sources matter for which question types in which geographies for which functions, and putting attention where it earns the most return. That's a different category of work than optimizing for any single platform — and it's the work most employer brand teams haven't started yet, because they didn't have visibility into the citation graph until very recently. (We've written elsewhere about why this is a measurement problem most enterprise teams haven't solved yet, and how to think about LinkedIn specifically as a ChatGPT input.)

That visibility now exists. The question is what to do with it.

Methodology

This analysis is based on more than 100,000 AI responses to employer-brand-related prompts captured between September 2025 and May 2026. Citations were extracted from the four AI engines that disclose their sources at response time: Perplexity, Google AI Overviews, OpenAI Search (with browsing enabled), and Google AI Mode.

The global ranking weights each employer in the dataset equally to avoid large-volume employers dominating the result. Presence rate is calculated per employer (the percentage of an employer's cited responses that include a given source) and averaged across employers. A volume-weighted version of this analysis produces the same top 10 with minor reordering of positions 4–8.

Prompts were tagged across four candidate-intent lenses — Discovery, Experiential, Competitive, Informational — to support the lens-level breakdown.

Top-10 rank ordering is high-confidence under both methodologies. Positions 15–20 sit within 0.1–0.3 percentage points of one another and should not be treated as strict ordinal rankings.