AI Visibility Report · May 26, 2026

Infento

A study of how generative search engines see this brand. Twenty-five category-relevant prompts run across two grounded LLMs and one training-memory LLM. Citations extracted, sources weighted, gaps surfaced.
46%
Gemini
3.5 Flash · grounded
11/24 prompts
36%
Claude
Haiku 4.5 · grounded
9/25 prompts
4%
Llama 4
Scout · training memory
1/25 prompts

TL;DR

Infento leads its competitive set on AI search (46% citation rate, ahead of LEGO Technic at 33%) but is invisible on the highest-intent buyer queries: gift, longevity, and "balance-bike-that-grows-with-the-child." Worse, the brand barely exists in LLM training memory (4% on ungrounded Llama), meaning offline-mode AI assistants treat it as if it doesn't exist.


The citation picture

Across 24 grounded Gemini queries about modular kids vehicles, build-it-yourself toys, STEM gifts, and longevity-positioned products:

Brand Citation rate
Infento 11/24 (46%)
LEGO Technic 8/24 (33%)
Berg Toys 3/24 (13%)
Strider 2/24 (8%)
Razor 2/24 (8%)
Kettler 0/24 (0%)

Infento leads the competitive set in AI search. The strength is concentrated in three themes: modular kids vehicles (3/3 cited), electric kids vehicles (2/2), sustainable toys (1/1). The weakness is concentrated in the themes that actually drive purchase decisions.


How different AIs see Infento

The three-LLM comparison surfaces the key strategic finding:

Model Citation rate What it measures
Gemini 3.5 Flash (grounded, live web search) 46% Current web presence
Claude Haiku 4.5 (grounded, web_search tool) 36% Current web presence via Anthropic
Llama 4 Scout (ungrounded, training only — no web access) 4% What an LLM "knows" from training data

The 37-percentage-point gap between grounded LLMs and Llama's training memory is the finding to act on. Infento's SEO and PR are working today: when an AI searches the web, it finds Infento often. But for an AI answering from training memory alone (offline-mode assistants, older model snapshots, devices without live search), Infento is functionally invisible. LEGO and Strider have decades of definitional content baked into LLM training corpora. Infento doesn't, yet.

This matters more every quarter as AI assistants on iOS/macOS/Android default to training-memory responses for speed and battery reasons.


Where Infento is losing

AI search skips Infento on three high-intent prompt categories:

1. "Modular balance bike that converts to a real bicycle"

2. "Premium STEM gift for a child who loves engineering"

3. "Toy that lasts from toddler to teenager"

These three gaps are not edge cases. They are the three highest-intent buyer queries in the entire test set (gift-buying, premium positioning, parent justification for spend).


What AI trusts in this category

The top-cited sources (excluding Infento's own domain) show how the AI grounds its answers:

Source Citations Role
youtube.com 58× Video demos, unboxings, family vlogs — heavy AI weight
reddit.com 21× Unfiltered parent discussions about toy durability and value
littlebigbikes.com Niche authority on kids' bikes — gatekeeper for balance-bike queries
rascalrides.com Reviews kids' ride-ons, especially electric
reviews.io Aggregated product reviews

The pattern: AI trusts community video (YouTube), peer discussion (Reddit), and a small set of niche-authority blogs. To win the gap queries above, Infento needs to be reviewed on Little Big Bikes and Rascal Rides — not just have content on infento.com.


One action this week

Pitch the top three niche authority sites — Little Big Bikes, Rascal Rides, and one parenting STEM publication of your choice — for dedicated reviews of the Infento Innovator kit framed around the exact phrases AI uses for the gap queries: "convertible balance bike," "premium STEM gift," and "toy that grows with the child from toddler to teenager." This directly seeds the grounding sources AI uses to answer the queries Infento currently misses.

In parallel, commission two long-form YouTube reviews (the platform AI cites most heavily at 58×) with the same semantic anchors in titles and descriptions. The combination targets both grounded LLMs (via the niche-blog reviews) and the longer training-memory tail (via YouTube content that future models will train on).