Overview
Stellagent launched Stella LLMO, an LLM optimization service for retailers and e-commerce businesses, on July 28, 2026. Stella LLMO helps a company's products appear as candidates when consumers ask generative AI for recommendations. Starting from real measurement of how a brand is seen inside AI, the service provides end-to-end support across seven areas: technical readiness of the site, product-data preparation, strengthening external information, distribution to AI channels, purchase pathways, and measurement. Alongside the launch, a free "AI Visibility Diagnosis" also began accepting applications the same day.
For details, see the press release on PR TIMES.
Stella LLMO measures how a brand and its competitors appear inside AI, then covers technical readiness, product data, distribution to AI channels, purchase pathways, and measurement
Background
The way people search for products is shifting from search engines to generative AI. While this improves convenience, it introduces two structures that search did not have. First, the candidates presented narrow from ten links to a handful of recommendations. Second, AI answers are shown not as a list of links but as definitive explanations, which consumers tend to accept as fact.
Our own measurement studies confirm both. In a survey asking AI about accommodations on Okinawa's main island, 84 properties appeared across 625 recommendation slots, but the top five alone accounted for 44.2%. Businesses that used to be found in the search long tail are not even entering the candidate set in AI answers. In a separate study asking five major AI services about electronics retailers' return and warranty policies, only 60 of 100 question patterns (60.0%) were answered correctly, while 25 (25.0%) were wrong — 15 of them in ways that could cause real harm.
The problem is that businesses cannot fix these issues on their own. AI answers cannot be edited directly, and the means to notice errors are limited. What a business can do is deliver information in a form that AI can find and correctly understand. Conversely, the businesses that have not organized their information are the ones that are disproportionately left undiscovered and mis-described. As a developer of agentic commerce infrastructure, Stellagent launched this service so that businesses can build a state in which AI finds and correctly understands their products.
Key Features
Start from measurement and identify the bottleneck before acting
We design roughly 20 questions by purchase scenario, run the same conditions repeatedly across multiple AIs, and compare mention rate, recommendation rate, citation status, and information accuracy for both the company and its competitors. Because generative AI answers vary each time even for the same question, we do not treat a single result as a ranking; we run repeatedly, record the distribution, and set baselines. We identify the losing questions, competitors, and causes before deciding the order of improvements, and we continue measuring monthly to keep up with model updates and information errors.
Before a product is recommended by AI, there are four gates: being read by AI, being understood as a product, being selected in comparison, and being buyable through AI. Measurement isolates which stage is blocking progress.
Not more articles, but a state AI can read
We inspect robots.txt, WAF, noindex, and JavaScript rendering to make important product pages discoverable and retrievable. We then unify product names, model numbers, prices, stock, and use cases, reflecting the same information across product pages, structured data, and product feeds. Outside the company's own site, we align product names and claims across press releases, comparison articles, videos, and retailer pages to reinforce the context and trust signals AI uses to judge. Measures discussed under LLM optimization tend to center on content production and technical fixes, but we define our scope as seven areas — from diagnosis to implementation, product-data distribution, purchase pathways, and measurement — and prioritize the work based on the diagnosis.
Don't stop at being found — connect through to purchase
As a developer of agentic commerce infrastructure, Stellagent has handled the full flow of AI finding a product, checking stock, and moving to purchase. Stella LLMO is designed to include product-feed connections to the shopping surfaces AI references, as well as product and inventory APIs and cart and payment integration. We track display, citation, AI-driven traffic, and purchase in a single metric system, and decide the next improvement target based on business results.
Note that we cannot manipulate the content of generative AI answers themselves, and we do not guarantee placement or recommendation ranking. This service prepares the conditions under which AI can correctly understand, compare, and recommend a product. Technical fixes may be reflected within a few weeks, but changes in recommendations and citations also depend on the accumulation of information, so we use three to six months as a guideline.
Service Overview
| Item | Details |
|---|---|
| Service name | Stella LLMO |
| Launch date | July 28, 2026 |
| Target | Retailers, e-commerce businesses, manufacturers |
| Scope | AI visibility assessment, crawl/index foundation, product and brand data foundation, external information and recommendation evidence, product-data distribution to AI channels, AI-based purchase experience, and measurement with ongoing improvement (seven areas) |
| AI services measured | ChatGPT, Gemini, Claude, Perplexity, Microsoft Copilot, and others |
| Initial fee | Fixed 200,000 yen per site (excl. tax), approximately four weeks |
| Ongoing operation | From 300,000 yen per month (excl. tax), six-month minimum |
| Service site | Stella LLMO service page |
| Free AI Visibility Diagnosis | Apply for the AI Visibility Diagnosis |
Looking Ahead
Stellagent will continue measuring how generative AI recommends products and purchase sources and will publish the results as survey reports. In areas where AI-driven purchasing leads to actual transactions, we will expand the range of product-data destinations and payment and order integrations, helping retailers and e-commerce businesses build a state in which they are discovered on AI, correctly understood, and connected through to purchase.