March 23, 202612 min read

Recommendations vs Mentions in AI: What Matters More for Brands — Data Across 5 Niches

AI can mention a brand and recommend it — these are different metrics. Recommendation rate analysis across 5 niches: travel, e-commerce, EdTech, hosting, FinTech. GEO Scout data from 9 providers.

Vladislav Puchkov

Founder of GEO Scout, GEO optimization expert

When a marketer sees their brand mentioned in 87% of ChatGPT responses, the first reaction is excitement. But behind that number hides a question that changes the entire picture: what exactly does AI say? Lists it among ten options — or directly recommends it to the user?

We analyzed GEO Scout data across the 5 largest niches in the Russian market — travel, e-commerce, EdTech, hosting, and FinTech — and found that the gap between mention rate and recommendation rate is massive. In some niches, AI mentions brands willingly but recommends only one in ten. In others — it recommends nobody at all.

This article is the first study of recommendation rate by niche in the Russian market.

Three Metrics You Shouldn't Confuse

Before diving into the data, let's break down three key AI visibility metrics that are often mixed up.

Mention Rate — "AI Knows About the Brand"

The percentage of AI responses where the brand is mentioned. If the brand appeared in 270 out of 300 responses — mention rate is 90%. The mention can be anything: in a list, comparison, or context. The brand simply appears in the response.

Positive Rate — "AI Speaks Well of It"

The percentage of mentions with positive tone. AI uses words like "convenient," "reliable," "high-quality," describes advantages. High positive rate means when AI talks about the brand — it talks positively.

Recommendation Rate — "AI Directly Suggests It"

The percentage of responses where AI explicitly recommends the brand to the user. Not just mentions in a list, but says: "I recommend," "best choice," "I suggest considering." This is the most valuable metric — a direct recommendation from AI converts.

Key distinction: a brand can have 92% mention rate, 92% positive rate, and only 4% recommendation rate. AI knows the brand, praises the brand, but doesn't recommend the brand. These are three fundamentally different levels of trust.

Travel: Why AI Boldly Recommends Aviasales

Travel is the niche with the highest recommendation rate among all studied industries. But "travel" is an understatement. More accurately: Aviasales is the only brand with mass recommendations.

Aviasales's numbers look anomalous against the entire market. A 36.6% recommendation rate means that in more than every third response, AI directly suggests Aviasales to the user. For comparison: the best e-commerce brand — Wildberries — has a 9.28% recommendation rate.

Which Providers Recommend Aviasales

Gemini recommends Aviasales in three out of four responses. Grok — in seven out of ten. Google AI Mode — in every other one. This isn't coincidence — it's a pattern.

Why Aviasales Specifically?

AI recommends Aviasales so actively for several reasons. First, low risk for the user: recommending a flight aggregator isn't recommending a specific airline. Aviasales doesn't sell tickets — it compares prices.

Second, no alternative with comparable recognition in the Russian-language segment. AI needs to recommend someone, and Aviasales is the obvious category leader.

Third, massive content trail: reviews, rankings, media mentions, feedback — AI is trained on a huge volume of positive content about this brand.

All other travel brands remain below 1% recommendation rate. AI knows about Tutu, Ostrovok, OneTwoTrip — but doesn't recommend them.

E-commerce: Praised But Not Recommended

E-commerce is the second niche by recommendation rate, but the picture is fundamentally different. Brands get high mention rate and positive rate, but recommendations stay modest.

Ozon: The Most Illustrative Case

Ozon perfectly illustrates the gap between "praising" and "recommending." Per GEO Scout data:

• Mention rate: 92.41% — AI knows about Ozon and mentions it almost always
• Positive rate: 92.24% — when it mentions it, it speaks well. "Convenient," "fast delivery," "wide assortment"
• Recommendation rate: 4.22% — but directly recommends only in 4% of cases

AI says: "Ozon is one of the largest marketplaces with a convenient interface." But doesn't say: "I recommend buying on Ozon." Praising doesn't mean recommending.

AliExpress: Zero Recommendations

AliExpress is mentioned in 52.74% of responses — AI knows about it. But recommendation rate is exactly 0%. No provider ever directly recommended AliExpress. Likely reason — reputational risks: slow delivery, product quality, questionable service. AI knows about these problems and won't stake its "reputation" on this recommendation.

EdTech: AI Is Cautious About Education

EdTech is a niche where AI shows notable caution. The leader's recommendation rate is just 7.14%.

Netology vs Skillbox: The Mention ≠ Recommendation Paradox

Netology and Skillbox have nearly identical mention rates: 86.55% and 86.13%. AI knows them equally well. But recommendation rate differs by 5.7x: 7.14% for Skillbox vs 1.26% for Netology.

Hexlet: Small But Recommended

Hexlet's mention rate is only 37.82% — AI knows it three times less than Skillbox. But recommendation rate is 4.62%, higher than Netology's. This confirms: a small niche brand with clear positioning can get more recommendations than a large generalist brand.

FinTech: AI Won't Take Responsibility for Banks

FinTech is the niche with the lowest recommendation rate of all studied. No bank crossed the 2.1% barrier.

Financial recommendations are a zone of maximum responsibility. If AI recommends a bank and the client loses money on unfavorable terms — it's a reputational risk for the AI system. LLM developers understand this and build caution into models.

Compare: Aviasales is recommended in 36.6% of responses (low user risk), while Alfa-Bank in 2.07% (high risk). A 17.7x gap — and it's no accident.

Hosting: Mid-Range Caution Zone

Hosting ranks fourth among niches by recommendation rate.

Timeweb is the absolute niche leader by awareness (89.80% mention rate) and the only hosting provider with a non-zero recommendation rate. But 2.45% is modest.

Summary Table: Recommendation Rate by Niche

The pattern is clear: the higher the potential damage from a wrong choice, the lower the recommendation rate. A bad flight ticket — a few hours lost. A bad bank — money lost. AI calibrates caution to the degree of responsibility.

Which Providers Recommend More Often

Not all AI providers behave the same. GEO Scout data reveals clear patterns.

Grok: The Most "Recommendation-Friendly" Provider

Grok consistently shows the highest recommendation rates across all niches:

• Aviasales: 70% recommendations
• Wildberries: 16.7%
• Skillbox: 16.7%

Grok isn't afraid to give direct recommendations. Its behavioral model is to help the user make a specific choice, not list all options.

Gemini: Close to Grok

Gemini is the second most active in recommendations:

• Aviasales: 76.7% (even higher than Grok)

Gemini is especially active in recommending travel brands. Its connection to the Google ecosystem may give it more confidence in service data.

ChatGPT: The Neutral Observer

ChatGPT almost never gives direct recommendations. Across all niches, its recommendation rate is near zero or well below average. ChatGPT's behavioral model is to list options, describe pros and cons, but not suggest a specific choice.

For brands, this is critical: if your GEO strategy is built on ChatGPT, recommendation rate will be low by definition, regardless of optimization quality.

Provider Summary

What Determines a Brand's Recommendation Rate

Analysis across 5 niches reveals factors that influence the shift from "AI knows" to "AI recommends."

1. Low Risk for the User

Aviasales (aggregator, not seller) — 36.6%. Alfa-Bank (financial decisions) — 2.07%. AI calibrates recommendations to the responsibility level.

2. Category Dominance

Aviasales is the obvious leader in ticket search. Wildberries is the first association with marketplace. When AI needs to recommend someone, it picks the brand with the strongest category association.

3. Clear Positioning

Hexlet (niche IT education) — 4.62% recommendation rate at 37.82% mention rate. Netology (broad profile) — 1.26% at 86.55%. The clearer the positioning, the easier it is for AI to formulate a recommendation.

4. Positive Content Trail

Brands with an array of positive reviews, rankings, and feedback on independent platforms receive more recommendations. AI is trained on this data and reproduces the expert community consensus.

5. Absence of Negative Signals

AliExpress — 0% recommendation rate at 52.74% mention. AI knows about delivery and quality problems. One strong negative signal can zero out recommendations despite high awareness.

What Brands Should Do: From Mention to Recommendation

If You Have High Mention Rate But Low Recommendation Rate

This is the Ozon situation (92% mention, 4% recommendation) or Netology (87% mention, 1.3% recommendation).

Actions:

1. Strengthen expert content — reviews with specific data, comparisons, case studies
2. Increase presence on ranking platforms — AI considers positions in rankings and catalogs
3. Work with reviews — a mass of positive reviews on independent platforms creates a "recommendation backdrop" for AI
4. Create content with clear recommendation language — if experts write "we recommend," AI is more likely to reproduce this pattern

If You're In a Low Recommendation Rate Niche

FinTech and hosting — niches where AI recommends almost nobody. That's not a brand problem but a niche characteristic.

Actions:

1. Focus on mention rate and position — if AI doesn't recommend banks, fighting for recommendation rate is futile
2. Improve tone — positive rate becomes the key metric in niches without recommendations
3. Work on domain citation — if AI won't recommend, at least get it to link to your site

Key Takeaways

Mention ≠ recommendation. Netology is mentioned in 86.55% of AI responses but recommended in 1.26%. Ozon — 92.41% mention, 4.22% recommendation. Positive rate also doesn't equal recommendation rate: Ozon has 92% positive mentions but is recommended in only 4%.

AI calibrates caution to responsibility. Travel — up to 36.6% recommendations (low risk). FinTech — up to 2.07% (high risk). This isn't a bug — it's a deliberate AI strategy.

Aviasales is a unique phenomenon. The only brand with mass recommendations across all providers. A ratio of 1 recommendation per 2.1 mentions — unattainable for other niches.

Providers differ radically. Grok and Gemini recommend brands 5-10x more often than ChatGPT. Multi-provider monitoring isn't optional — it's necessary for understanding the real picture.

Recommendation rate is the conversion metric for AI. Mention rate is reach. Recommendation rate is conversion. GEO strategy focus should shift from "how to get into the response" to "how to get a recommendation."

Full data by brand and provider available in the interactive ranking at geoscout.pro. The Command Center of GEO Scout turns monitoring data into a concrete action plan — what steps to take so AI doesn't just mention but recommends your brand.