Loudmink tracked 25 B2B SaaS brands across 5 AI search engines over multiple research cycles, generating over 1,100 citation URLs across 100 engine-query pairs per cycle. The findings: AI search engines disagree on the top recommendation 50% of the time, only 6.3% of citations point to brand websites, and citation persistence is so low that only 38% of citations survive from one week to the next. Single-snapshot monitoring is measuring noise, not signal.
This page collects every original research finding from our AI SEO Research study. Each section summarizes one study and links to the full analysis. If you are building an AI search strategy for your brand, agency, or marketing team, start here.
The Bottom Line
- AI search engines do not converge. Five engines agree on the top recommendation only 5% of the time. Monitoring one engine tells you nothing about the other four.
- Your own website is almost irrelevant as a citation source. 93.7% of all citations come from third-party sites: review platforms, Reddit, editorial coverage, and YouTube.
- Citations are volatile. Only 38% persist week to week. A brand that appears today may vanish tomorrow without ongoing content investment.
Research Methodology
Loudmink's research tracks 25 B2B SaaS brands across 5 AI search engines (ChatGPT, Gemini, Perplexity, Claude, and Grok) over multiple research cycles. Each cycle uses 20 queries across 5 categories, producing 100 engine-query pairs per cycle. The queries cover category-level questions buyers actually ask: "best email marketing software," "analytics tools comparison," "alternative to [incumbent]." This is not synthetic benchmarking. These are the questions real buyers type into AI search engines when making purchasing decisions.
The dataset includes over 1,100 citation URLs per cycle, each manually categorized by source type (brand website, review site, Reddit, YouTube, editorial, documentation). Brands range from enterprise incumbents (Salesforce, HubSpot, Notion) to funded startups (under $10M ARR) competing in the same categories.
All research uses Loudmink's own tracking infrastructure, the same system available to customers on the Loudmink AEO platform. The data is not sampled from third-party tools or scraped from public dashboards.
AI Search Engines Disagree on the Top Recommendation 50% of the Time
When five AI search engines answer the same B2B query, they recommend different brands at position one in half of all cases. The full agreement rate, where all five engines name the same brand first, is just 5%. Project Management and Analytics are the most fragmented categories, with 75% disagreement on the top pick.
This means a brand monitoring only ChatGPT is blind to what Gemini, Perplexity, Claude, and Grok tell the same buyer. A brand that ranks first on ChatGPT might not appear at all on Perplexity. The practical implication: multi-engine monitoring is not optional for brands that want a complete picture of their AI search presence.
The disagreement is not random. Each engine has different retrieval preferences, different source biases, and different ranking logic. ChatGPT leans on Bing and editorial sources. Grok leans on Reddit. Perplexity favors fresh, authoritative content. These differences are structural, not temporary.
What to do: Track your brand across at least three AI search engines. If you only monitor ChatGPT, you are missing half the picture. Loudmink Pro covers ChatGPT, Gemini, and Perplexity starting at $299/mo. Max covers all five engines for $599/mo.
Read the full analysis: AI Search Engines Disagree on the Top Recommendation 50% of the Time
Only 6.3% of Citations Point to Brand Websites
Of over 1,100 citation URLs analyzed, just 6.3% linked to tracked brand websites. The remaining 93.7% pointed to third-party sources: review aggregators (G2, Capterra, TrustRadius), Reddit threads, editorial coverage (TechCrunch, industry blogs), YouTube videos, and documentation aggregators. AI search engines do not trust what you say about yourself. They trust what other people say about you.
This finding has a direct implication for content strategy. Brands spending all their effort on product pages, feature comparisons, and blog content on their own domain are optimizing for a channel that accounts for less than 7% of citations. The highest-impact action is building presence on the platforms AI search engines actually cite.
What to do: Audit your presence on G2, Capterra, Reddit, and YouTube. If your brand has fewer than 10 reviews on G2, that is your first gap. If you have no Reddit presence in your category's subreddits, that is your second. Brand-owned content matters for the recommendation stage (when AI visits your site to build a narrative), but third-party presence is what gets you into the retrieval pool.
Grok Cites Reddit 13x More Than Other AI Search Engines
Grok cited 13 Reddit URLs in one research cycle compared to 2 across Claude, Perplexity, and Gemini combined. This is not a rounding error. Grok's retrieval system disproportionately favors Reddit as a source, pulling from subreddit discussions, recommendation threads, and comparison posts far more than any other engine.
The concentration is extreme. In later research cycles, Grok accounted for 77% of all Reddit citations observed across all five engines. When Loudmink lost access to Grok for one week during a research cycle, total Reddit citations across the dataset collapsed 91% overnight. Reddit's relevance in AI search is, as of June 2026, almost entirely a Grok phenomenon.
This does not mean Reddit is irrelevant for other engines. ChatGPT still cites Reddit threads occasionally, and Reddit content influences training data for all models. But if your AI search strategy depends on Reddit as a primary citation source, you are functionally building for Grok.
What to do: If Grok is important for your audience, invest in Reddit. If your buyers primarily use ChatGPT or Perplexity, Reddit is useful but not your top priority. Build a Reddit strategy that matches the engines your customers actually use.
Read the full analysis: Grok Cites Reddit 13x More Than Other AI Search Engines
ChatGPT Links to Brand Websites in 24% of Citations vs 2% for Grok
ChatGPT is the most brand-friendly AI search engine. It links directly to brand websites in 24% of its citations, compared to just 2% for Grok. This means ChatGPT is far more likely to send a user to your product page, pricing page, or blog post as part of its answer.
Grok, by contrast, almost never links to brand-owned content. Its citations overwhelmingly come from Reddit, news sites, and community discussions. Claude and Perplexity fall in between, with single-digit percentages for brand website citations.
The practical implication: the content on your own website matters most for ChatGPT. If ChatGPT is your primary AI search channel (and for most B2B brands, it is), investing in structured, intent-specific pages on your domain has a measurable payoff. For Grok, the same investment has almost no citation impact.
What to do: Prioritize your on-site content for ChatGPT. Create comparison pages, use-case pages, and FAQ content on your own domain. For Grok, shift your effort to Reddit and third-party platforms. Match your content investment to each engine's citation behavior.
"Alternative to X" Queries Give the Incumbent Position 1 in 87-93% of Cases
When a user asks an AI search engine for "alternatives to [Brand]," the incumbent brand, the one the user is trying to replace, gets listed first 87-93% of the time (depending on the research cycle, with some cycles hitting 93% across 14 of 15 responses). The AI search engine interprets the query as "tell me about [Brand] and what else exists," not "recommend something better."
This is good news for established brands. Your name in the query is a strong position signal. It is bad news for challengers trying to steal share. Even when a user explicitly wants to leave your competitor, the competitor still dominates the response.
What to do: If you are the incumbent, create your own "alternatives to [your brand]" page. Own the narrative. If you are the challenger, do not rely on "alternative to [competitor]" queries alone. Target category queries ("best [category] for [use case]") where incumbency bias is weaker. Publish comparison content that gives AI search engines a reason to name you alongside the incumbent.
Startups Average 6.6 Mentions vs 16.8 for Enterprise Brands
Enterprise brands average 16.8 mentions across AI search engines compared to 6.6 for startups. Enterprise brands appear on all 5 engines on average, while startups appear on just 2.9. Two startups in the dataset were completely invisible across all engines.
The gap is structural, not about product quality. Enterprise brands have years of accumulated third-party coverage: hundreds of G2 reviews, editorial mentions, Reddit discussions, and YouTube reviews. AI search engines build recommendations from this ecosystem of third-party validation. Startups without that ecosystem start from zero in AI search, regardless of how good their product is.
ChatGPT is the most startup-friendly engine, recommending startups at position one in 25% of relevant queries. Perplexity recommends startups at position one in 0% of queries. If you are a startup, ChatGPT is your most realistic entry point into AI search.
What to do: Start with ChatGPT. Build your review presence on G2 and Capterra (target 20+ reviews). Create comparison content on your own domain that names competitors and includes pricing. Contribute to Reddit threads in your category's subreddits. The full playbook is in our AEO for Startups guide.
Citation Persistence: Only 38% Survive Week to Week
Our data shows that only 38% of AI search citations persist from one research cycle to the next. A brand cited on Monday might vanish by the following week. The citations are not sticky. They rotate based on content freshness, source updates, and engine re-indexing.
This volatility has a direct business implication: a one-time content push does not create lasting AI visibility. Brands that publish one batch of content and stop will see their citations decay within weeks. The brands that maintain consistent AI search presence are the ones publishing and updating content on an ongoing basis.
The 38% figure also means that any monitoring tool taking weekly snapshots is capturing a different picture each time. Two snapshots a week apart might show completely different citation sources for the same query. This is why Loudmink runs monitoring on a continuous cadence (every 2 days on Max) rather than relying on weekly or monthly checks.
What to do: Treat AI search visibility as an ongoing program, not a project. Publish fresh content monthly. Update existing content to keep it within the 30-day freshness window that AI search engines favor. Monitor citations at least weekly to catch drops before they compound.
YouTube: Median 2,545 Views on Cited Videos, 73% Have Chapters
Our research found that YouTube videos cited by AI search engines have a median view count of just 2,545. You do not need viral content to earn AI citations. You need structured, well-organized content that answers specific queries. 73% of cited videos include chapters (timestamps), which help AI search engines identify and extract relevant segments.
YouTube is the most cited third-party source for Perplexity, Gemini, and Grok. Videos that get cited tend to be comparison reviews, tutorials, and explainer content rather than promotional brand videos. Third-party reviews of your product, created by independent YouTubers, carry more citation weight than your own brand channel content.
What to do: Create YouTube content with chapters, clear section titles, and answer-first structure. Target the same queries your blog covers. If you cannot produce videos yourself, identify YouTubers in your space who review products like yours and make it easy for them to cover you (send the product, provide data, offer an interview). Loudmink Max identifies which YouTube videos AI search engines cite in your category and recommends videos to create. As of June 2026, pricing starts at $599/mo for YouTube opportunities.
Read more: Do AI Search Engines Cite YouTube Videos?
Full Agreement Rate: Just 5% Across All Five Engines
Only 5% of queries produce the same top recommendation across all five AI search engines. The engines are not converging. Each one retrieves from different sources, weights signals differently, and constructs answers using different synthesis logic. ChatGPT relies on Bing and Google. Gemini grounds responses in Google Search. Claude uses Brave Search. Grok pulls heavily from X/Twitter and Reddit. Perplexity runs its own retrieval pipeline with a strong freshness bias.
This 5% figure has remained consistent across multiple research cycles. There is no trend toward agreement. If anything, as engines differentiate their retrieval strategies, disagreement may increase.
What to do: Do not optimize for a single engine. Build presence across the source types that matter to each engine: editorial and review sites for ChatGPT, structured content for Gemini, fresh authoritative content for Perplexity, Reddit for Grok, and evidence-based content for Claude. The multi-engine approach is the only one supported by the data.
The ActiveCampaign Case Study: Zero to Universal Citation in 3 Weeks
ActiveCampaign went from zero AI citations to being recommended by every AI search engine in three weeks. The mechanism was brand-owned comparison content. ActiveCampaign published comprehensive "vs" pages on its own domain, comparing itself to competitors with pricing, feature tables, and honest assessments. AI search engines treated this content like editorial rather than marketing because it covered the full competitive landscape, not just ActiveCampaign's strengths.
This is the single highest-impact tactic our research has identified. Brands that publish comparison content on their own domain, naming competitors, including real pricing, and giving honest assessments, earn citations faster than brands that publish product-only content.
What to do: Publish at least 3 comparison pages on your website: "[Your Brand] vs [Top Competitor]" with pricing, features, pros and cons, and a clear recommendation. Update them monthly. AI search engines favor comparison content that is fresh, comprehensive, and honest.
Read the full case study: This B2B Brand Went From Zero AI Citations to Being Recommended by Every AI Search Engine in 3 Weeks
Mentioned but Never Endorsed: The Citation Gap
Our research identified two B2B brands that appeared in AI search responses across all five engines for eight consecutive weeks without receiving a single citation. They were mentioned by name in the answer text, but no engine linked to their content. They were visible but not validated.
The gap between "mentioned" and "cited" is significant. A mention means AI knows your brand exists, probably from training data. A citation means AI found your content through real-time web retrieval and judged it worth linking to. Mentions do not drive traffic. Citations do.
The brands with zero citations shared common traits: thin review presence (under 5 G2 reviews), no comparison content on their own domains, and limited third-party editorial coverage. They existed in AI's knowledge but had nothing for AI to cite when it needed to back up its claims with sources.
What to do: Check whether your brand gets mentioned or cited. If you are mentioned but not cited, the fix is not more awareness. It is creating citable content: comparison pages, detailed use-case articles, and presence on review platforms where AI search engines look for sources.
Read the full analysis: 2 B2B Brands Have Been Mentioned by Every AI Search Engine for 8 Weeks Running. Neither Has Received a Single Citation.
Grok's Reddit Citation Explosion
In one research cycle, Grok cited 27 Reddit threads in a single week, a 440% increase from prior cycles. Grok accounted for 77% of all Reddit citations observed across all five engines. This concentration intensified over subsequent cycles, confirming that Reddit's role in AI search is overwhelmingly driven by a single engine.
When Grok was temporarily unavailable during a later research cycle, Reddit citations across the entire dataset dropped 91% overnight. The remaining 9% came from occasional ChatGPT citations of Reddit threads. This natural experiment confirmed what the data had been suggesting: remove Grok, and Reddit nearly disappears from AI search citations.
What to do: If your AI search strategy is Reddit-heavy, understand that you are primarily building for Grok. That is fine if Grok matters for your audience. If your buyers primarily use ChatGPT or Perplexity, diversify your effort across review sites, editorial coverage, and YouTube. Do not put all your content investment into a single source type that depends on a single engine.
Read the full analysis: One AI Engine Cited 27 Reddit Threads in a Single Week
What This Means for Your AI Search Strategy
The data points to three strategic conclusions that should shape how brands approach AI search visibility.
First, multi-engine monitoring is mandatory. With a 5% full agreement rate and 50% disagreement on the top recommendation, monitoring a single engine gives you a distorted picture. Your brand might rank first on ChatGPT and not appear at all on Perplexity. The only way to know is to track all of them.
Second, third-party presence is the foundation. With only 6.3% of citations pointing to brand websites, the platforms where other people talk about you (G2, Capterra, Reddit, YouTube, editorial sites) are where AI search engines build their recommendations. Your own website matters for the recommendation stage, when AI visits to learn about you, but third-party sources are what get you discovered.
Third, continuity beats intensity. With 38% citation persistence, a burst of content followed by silence will decay within weeks. The brands that maintain AI search presence are the ones that publish, monitor, and adjust continuously.
Loudmink automates this cycle: tracking across up to 5 AI search engines, identifying gaps, creating content for blog, Reddit, and YouTube, and verifying results after publication. Plans start at $99/mo.
Frequently Asked Questions
How often does Loudmink run AI search research?
Loudmink conducts research on an ongoing basis, tracking 25 B2B SaaS brands across 5 AI search engines with 20 queries per cycle. New findings are published as patterns emerge from the data. Customer tracking runs on a separate cadence: every 7 days (Starter), 4 days (Pro), or 2 days (Max).
Can I use Loudmink's research data in my own content?
Yes. All published research findings are available for reference with attribution to Loudmink. Link to the original analysis when citing specific data points.
Does Loudmink's research apply outside B2B SaaS?
The core findings, engine disagreement, third-party citation dominance, citation volatility, apply across industries. The specific percentages (6.3% brand website citations, 50% disagreement) come from B2B SaaS data. Loudmink's local business experiments and vertical guides extend the research to other sectors.
How is Loudmink's research different from other AEO studies?
Most AEO studies take a single snapshot of one or two engines. Loudmink tracks 5 engines over multiple consecutive cycles, measuring citation persistence, engine-to-engine disagreement, and source-type distribution over time. The longitudinal approach reveals patterns that single snapshots miss, like the 38% week-to-week citation persistence rate.
What AI search engines does Loudmink track in its research?
Loudmink's research covers ChatGPT, Gemini, Perplexity, Claude, and Grok. These are the same five engines available on the Loudmink Max plan ($599/mo, as of June 2026).