Which sources do LLMs cite when buyers search for software? The data from 20,160 chats

Key takeaways

• Around 81% of review platform citations go to multi-vendor pages, not individual profiles. Comparisons, listicles, and category pages drive the vast majority of retrievals.
• OMR Reviews is the #1 cited single source across all 8,035 domains analyzed, ahead of YouTube, Reddit, and every vendor website in the study. (Yes, we ran this study together with OMR Reviews. Make of that what you will, the data doesn't care.)
• Vendor sites dominate as a source type but are fragmented across hundreds of individual domains. Review platforms concentrate on a handful of addresses, making them the strongest individual sources.
• The model your buyers use changes the picture significantly. Copilot is the most review-heavy at 9.9%; Gemini cites almost no external sources at all.
• Category determines how much review platforms matter. IT Service Management (11.6%) and HR Software (9.6%) are review-platform-heavy; E-Mail-Marketing (3.8%) and SEO tools (3.9%) are not.
• Three levers drive AI visibility in software search: (1) building comparison and alternatives content on your own website, (2) getting featured in the right lists on review platforms, and (3) diversifying into editorial and UGC sources.

Software buying decisions are shifting toward AI. When buyers ask ChatGPT, Copilot, or Claude which tool to pick, the AI pulls from a set of sources, and those sources shape the consideration set before any vendor website is visited. Our hypothesis going in: when users ask LLMs with clear commercial intent, review platforms would show up disproportionately often as sources. We wanted to know whether that was actually true, and, if so, which review platforms and page types appeared most often, and how the results varied across different AI models.

How we ran the study

This is a descriptive source analysis conducted by Radyant and OMR Reviews. The numbers come directly from aggregated Peec AI exports.

Here's the setup:

Each category's 12 prompts follow the same validated split we use across our AI Search tracking: 8 long-tail prompts that mirror how a real buyer actually phrases a question ("which tool handles X for a team of Y people"), 2 short-tail prompts that cover the generic head term a category gets searched under ("best CRM software"), and 2 comparison prompts that name two specific tools and ask the model to pick between them. We cover this distribution, and why it holds up better than generic keyword lists, in our guide to prompt tracking.

For each category, we chose two tools that are comparable in terms of product, performance, and service, but differ in their presence on OMR Reviews: one has a strong profile, while the other has little or no presence. This setup allows us to examine the impact of a tool's presence on OMR Reviews and review platforms more broadly on an LLM's answer.

Here's what the prompts actually look like inside Peec AI. A short excerpt from the IT Service Management category, including the deeploi vs. Atera comparison prompt:

All prompts were German-language and DACH-market focused. That context matters for interpreting the results, particularly OMR Reviews' ranking, which reflects its dominant position in the German-speaking market. In English-language markets, G2 and Capterra would likely place higher.

What makes this data set meaningful is the prompt design: these aren't generic queries. They mirror the actual buying-intent questions real users ask LLMs when evaluating software: this is a signal worth measuring.

OMR Reviews comes out as the most-cited single source

Across all 8,035 unique domains analyzed in the study, one source came out on top: OMR Reviews, with a visibility score of 13.5%.

To put that in context:

OMR Reviews appears in more than three times as many AI answers as the strongest individual vendor website. That's a significant concentration of citation power in a single third-party domain.

Download PNGSVGShare on LinkedInCopy link

But here's the nuance that makes the finding more interesting and more useful than the headline number alone: when you look at source types rather than individual domains, vendor and manufacturer websites still dominate. They account for 62.5% of all citations. Review platforms, as a type, sit at just 6.3%. How can review platforms be both dominant and marginal at the same time?

The answer is distribution. Vendor citations are spread across hundreds of individual domains. Every software company in every category has its own website. Review platform citations, by contrast, concentrate on a handful of addresses. OMR Reviews, G2, Capterra, GetApp: a small number of platforms absorb a disproportionate share of all review-type citations.

But OMR Reviews doesn't lead on every model

The overall ranking tells one story. The model-by-model breakdown tells another, and it's one that software brands need to understand before drawing tactical conclusions.

Two findings here deserve special attention.

Gemini cites almost no external sources for these prompt types at all. The few hits it returns point to a Gemini-internal link. If your buyers use Gemini to research software, source optimization has almost no leverage there, at least for now.

ChatGPT is not particularly review-heavy. It's media-heavy: editorial sources and media sites account for 14% of ChatGPT citations, the highest share of any model. If you want visibility on ChatGPT specifically, media coverage, editorial mentions, and press placements matter more than review platform presence.

It also varies by category

The aggregate numbers are useful for orientation. But whether review platforms are a high-leverage channel for your software category depends significantly on which category you're in.

Review platform share of all citations, by category:

ITSM and HR Software show the highest review platform concentration, because multiple platforms are simultaneously active there with substantial visibility (in ITSM: Gartner 14%, Capterra 12%, G2 11%, OMR Reviews 10%).

One important nuance: OMR Reviews as an individual platform is actually strongest in Collaboration, SEO Tools, and Time Tracking (16 to 17% visibility), some of its best results in the study. But in those categories OMR Reviews dominates while other platforms are comparatively weak, which is why the aggregate type share still looks low. In SEO Tools specifically, that 3.9% shouldn't read as "review platforms don't matter": one platform is carrying the category almost single-handedly, making it one of the strongest single-platform opportunities in the study.

explore the data

Pick a model, a category, or a page type to see how much review platforms are cited as a source.

By modelBy categoryBy page type

review-platform share of all citations

Microsoft Copilot9.9%Claude6.7%Google AI Overview6.2%ChatGPT5.3%Gemini0.2%

Microsoft Copilot

The most review-heavy model. Capterra (9.3%) leads OMR Reviews (5%).

The most important finding: it's not the profile. It's the list.

This is the finding with the biggest strategic implications, and one that many software brands haven't fully acted on yet.

When we looked at which types of pages within review platforms were actually being cited by LLMs, the breakdown was stark:

Around 81% of all review platform retrievals go to multi-vendor pages: comparisons, listicles, and category pages. Only about 17% go to individual vendor profiles.

Download PNGSVGShare on LinkedInCopy link

That said, the profile is the foundation: to appear in OMR Reviews' category pages and listicles, you need a profile in the first place. And to increase your chances of being featured prominently in those lists, the number and quality of your reviews matter. More reviews signal credibility to the platform's ranking logic. The path to AI visibility on OMR Reviews runs through the profile, but it doesn't stop there.

The dominant page type varies by platform:

The strategic question isn't "do I have a profile on OMR Reviews?", it's "am I featured in the listicles and category pages on OMR Reviews?". And on G2, Capterra, and GetApp: "do I appear in the comparison and alternatives pages for my category?"

This also has direct implications for your own website. The same logic that makes LLMs prefer multi-vendor pages on review platforms applies to the open web: comparison and alternatives content (pages that put multiple tools in context and help a buyer decide) are the format AI models reach for when answering commercial software queries.

Building that content on your own domain is the one lever you control entirely.

What this looks like in practice: the deeploi case

The deeploi case makes the study findings tangible with real numbers.

deeploi is an IT management platform for growing companies. Radyant has been working with deeploi since November 2025. The strategy was built directly around the logic this study later confirmed: build comparison and alternatives content on the website, and establish a strong presence in the right places on OMR Reviews.

What was built:

• 40+ content pieces published on deeploi's website, predominantly comparisons, listicles, and alternatives articles (e.g. "Die 7 besten IT-Asset-Management-Softwarelösungen 2026", "Die 5 besten Atera Alternativen 2026 im Vergleich")
• Active OMR Reviews presence across all relevant categories, including IT Service Management, IT Asset Management, and Offboarding
• Consistent profile maintenance: complete information, content aligned with the website, ongoing review acquisition
• Multiple badges earned: 3× Leader (IT Service Management, IT Asset Management, Offboarding), 1× Top Rated (IT Service Management) for Q3/26

The AI visibility results:

• Brand visibility in AI Search (how often deeploi is named): from 0% at project start to an average of 15%, with a peak of 24.3%
• OMR Reviews retrieved share for deeploi (how often OMR Reviews is cited as a source): 14.4%, ranking #3 among all sources, behind only deeploi's own website and YouTube
• In comparison prompts (deeploi vs. Atera): LLMs pulled deeploi pages 32 times, Atera pages only 4 times

The business results:

• Organic web search leads: more than tripled (~+250%) since the partnership began.
• AI-attributed leads (self-reported): a new channel that didn't exist before the partnership.

A few important caveats: this is not a controlled experiment. Lead growth reflects multiple levers working in combination: content production, programmatic clusters, review platform presence, and ongoing AI visibility work. We cannot isolate the contribution of any single factor. The deeploi vs. Atera prompt result is a signal, not a causal proof.

What the case does show: when a software brand builds presence on exactly the page types and source categories that LLMs prefer (its own comparison content, OMR Reviews category pages and listicles, a well-maintained review profile), AI visibility follows, and that visibility can show up as measurable demand in the CRM.

What software brands should do next

The study points to a three-lever framework for AI Search visibility. None of the three levers work in isolation. The data consistently shows that the brands with strong AI visibility have all three working together.

Lever 1: Own your comparison and alternatives content

Build dedicated comparison pages on your own website. Tool A vs. Tool B. The top alternatives to Tool X. The best software options for use case Y.

These are the page types LLMs reach for most often when answering commercial software queries, and they're the ones you control completely. Your own domain is the only source where you can shape the narrative, the framing, and the specifics entirely. Vendor sites as a type account for 62.5% of all citations. Make sure your site is earning that share, not diluting it.

Lever 2: Get into the lists, not just the profile

Getting featured in the right lists starts with the profile, but it doesn't end there. Here's the path:

Step 1: Get listed. A free profile is enough to get started on OMR Reviews, no paid plan required. This is table stakes: without a profile, you can't appear in category pages or listicles at all.

Step 2: Get reviewed. The number and quality of customer reviews is what actually feeds the ranking logic behind category listicles. This works on a free profile just as it does on a paid one, though a paid profile can give additional visibility support on top. If you only do one thing after setting up your profile, prioritize collecting genuine customer reviews.

Step 3: Get featured. Profile and reviews are the prerequisites; the actual AI-visibility lever is showing up in the pages that get cited:

• On OMR Reviews: target inclusion in category listicles and category pages. In all 8 categories in our study, the most-cited OMR Reviews page was the listicle, not a vendor profile.
• On G2, Capterra, GetApp: focus on comparison and alternatives pages. That's where ~60% or more of their citations originate.

Make sure you appear in every relevant category, not just your primary one.

Lever 3: Diversify into editorial and UGC

YouTube (10.3%) and Reddit (8.6%) are the #2 and #3 most-cited sources overall. ChatGPT in particular is media-heavy: editorial sources account for 14% of its citations, the highest share of any model in the study.

A strategy that only optimizes for review platforms will have blind spots, particularly on ChatGPT. Editorial coverage, media mentions, and presence in relevant community discussions (Reddit threads, YouTube reviews) build a source mix that's more resilient across models.

One model to deprioritize for now: Gemini. In our data, Gemini cited almost no external sources for commercial software prompts. Until that changes, building source strategy specifically for Gemini visibility isn't a productive use of resources.

Limitations worth knowing

A few methodological points worth keeping in mind when reading these results:

Snapshot in time. These numbers reflect a four-week window in June 2026. LLM behavior, training data, and brand content all shift over time. The directional findings are likely to hold; specific percentages will move as models update and the competitive landscape changes.

DACH market scope. The prompt set is German-language and DACH-focused. OMR Reviews' leading position is partly a language and market effect. In English-language markets, international platforms like G2 and Capterra would likely perform stronger. Results apply to the German-speaking market.

Brand-agnostic prompts. Prompts are deliberately brand-neutral (buyer perspective, no brand names mentioned) to measure organic discovery rather than confirmation search. This structurally favors vendor/product sites as a type (62.5%), because LLMs tend to recommend vendor websites directly in response to unspecific buyer prompts. In prompts that already include a brand name, the model is more likely to validate than discover, which would likely surface more review and comparison content.

The bottom line

The Software Prompt Study points to a clear signal: when buyers ask LLMs for software recommendations, AI models cite concentrated third-party sources far more readily than individual vendor websites. Within those third-party sources, review platforms are the most powerful single-source category. And within review platforms, it's the multi-vendor pages (lists, comparisons, category pages) that drive the overwhelming majority of citations.

The brands showing up in AI answers have built all three levers: their own comparison content, presence in the right lists on the right review platforms, and a diversified source footprint across editorial and UGC. That combination is harder to build than a review profile, and considerably harder for competitors to copy.

This is what we measure, build, and optimize for at Radyant, for software brands serious about AI Search visibility.

FAQ