Parallel World
: How to use it, features, and the business problems it solves
Add bookmark
What is Parallel World?
Parallel World is a B2B SaaS for LLMO/GEO that measures and monitors how your website looks to generative AI and AI agents across five stages: training, search indexing, retrieval at answer time, browsing and interaction, and transactions. Instead of inventing its own score, it records public standards from Cloudflare, Google, OpenAI, Anthropic and others, along with the results of each provider's official tools, and reports conflicts and changes between standards with sources in a Japanese PDF report. Scans are read-only. A free scan is available with no sign-up, and weekly monitoring costs ¥32,780 per site per month (tax incl.), paid by invoice. Operated by Migurei, a subsidiary of HBLAB Japan Inc.
Business problems it solves
About "Parallel World"
What is Parallel World?
Parallel World is a B2B SaaS for LLMO/GEO that measures and monitors, against published standards, how your website looks to generative AI and AI agents. Across the five stages at which AI touches a site — from training crawlers to agents that make purchases on a customer's behalf — it shows what AI can and cannot read. Results come as a Japanese PDF report.
It is run by Migurei, a subsidiary of HBLAB Japan Inc. (株式会社エイチビーラボジャパン, Minato, Tokyo), which is also the seller and the issuer of invoices. The official website states plainly that it is a measurement service, not an agency that carries out fixes or provides consulting.
training to transactions
is checked in robots.txt
sends per scan
no sign-up or card
Figures from the official website and documentation (as of October 2026).
Source: Parallel World official website (applies to the images below as well)
People see the page. AI reads the text
The website's example is a fictional coffee-bean product page. A person sees the product photo, the price "¥1,800" and "In stock", but AI retrieval tools discard the design and keep only the text. If the price is inside the photo and the stock status is rendered by a script, the two things customers most want to ask about never reach the AI. Parallel World lists, item by item, the gap between what people see and what AI receives.
Measuring the "five entrances" AI uses to reach your site
The website explains that the first entrances decide "whether AI knows you exist", and the later ones "whether AI can act on your customer's behalf".
| Stage | Who touches the site | What is checked | Availability (October 2026) |
|---|---|---|---|
| A. Training | GPTBot, ClaudeBot, Google-Extended, etc. | robots.txt, Content-Signal | Available |
| B. Search indexing | Googlebot, Bingbot, OAI-SearchBot, etc. | Sitemaps, structured data | Available |
| C. Retrieval at answer time | ChatGPT-User, Claude-User, etc. | Whether raw HTML contains price, stock, contact details, opening hours | Available |
| D. Browsing and interaction | Browser-based agents | Accessibility tree, form attributes | Basic items only (observing real assistants is Phase 2) |
| E. Transactions | Shopping agents | Whether payment methods are declared in machine-readable form (e-commerce only) | Phase 2 |
The scan checks robots.txt (access for about 15 major AI bots), llms.txt (presence and format), sitemap consistency (noindex, redirects, 404s), structured data per page type (Product, Offer, Organization, etc.), price, stock, shipping and return terms in the raw HTML, responses to published AI fetcher User-Agents (403, 429, challenges), and the accessibility tree and form autocomplete attributes (documentation, "What we check").
No proprietary score — just sources and where standards disagree
Some LLMO/GEO tools assign points using their own metrics. Parallel World creates no scoring criteria of its own; it records the standards published by Cloudflare, Google, OpenAI, Anthropic and others, along with the results of each provider's official tools and APIs, as they are. Results are expressed not as a total score but as "the number of published criteria you meet".
- Conflicts between standards, with sources — when standard A and standard B ask for different things, the conflict itself is shown as an item
- Past results are never rewritten when a standard changes — results measured under an old standard remain, labeled "measured under previous standard"
- Source and version for every item, plus a snapshot at measurement time — with a fix list you can hand straight to your web team
In the sample report, a fictional site shows the number of criteria met, such as "18/24 cleared", the counts of warnings, critical issues and items not applicable, and the score for each of the five entrances (A. Access for training bots 5/5, C. Facts readable at answer time 4/7, and so on). Critical items come with the standard they are based on, for example "Price is not in the text (source: Google structured data Product / Offer)".
A read-only scanner that won't break your site
The scanner identifies itself as migurei-ParallelWeb, and its requests carry Web Bot Auth (HTTP Message Signatures) signatures. If you disallow this token in robots.txt, the scan records "blocked" and ends without fetching content. It never submits forms, adds to carts, logs in or places orders; it sends at most 50 requests of its own per scan, at least 2–3 seconds apart for the same domain (documentation, "Scanner").
How to use
- Enter your domain in the free scan Enter the top-level domain name (e.g. yourcompany.co.jp) into the free scan on the official website, and choose whether it is your own site, a client's site or other (such as competitor research). No account or card is needed.
- Check the results A summary appears on screen within a few minutes. If you enter an email address, detailed results for each item are emailed once (the address is deleted right after sending). The results link is valid for 7 days and can be opened by anyone with the URL.
- Optionally receive re-scans for four weeks You can opt in to automatic weekly re-scans for four weeks (up to three times), with an email only when the results change.
- Apply for monitoring by email Send an email to contact@migurei.ai with the subject "お申込み（モニタリング）" (application form in preparation). A representative will contact you about the procedure and invoice.
- Follow up on fixes with the weekly report Hand the per-entrance reports (Japanese PDF) and fix lists to your web team and work through them.
Features
01Available now
- Free scan — check robots.txt, llms.txt, sitemaps and Google PageSpeed Insights results as retrieved
- Weekly monitoring — reports per entrance (Japanese PDF) and fix lists
- Detection of conflicts between standards — with hints on how to fix each, based on the source documents
- Public information for AI agents — llms.txt, Markdown versions of every page (Accept: text/markdown) and a read-only MCP server
02Public tools anyone can use
Three tools are published via the MCP server (https://pw.migurei.ai/mcp) and plain URLs.
- check_ai_bot_access — reads a domain's robots.txt once and returns access per AI bot (also available via URL without MCP)
- verify_ai_bot — given an IP address and User-Agent from your logs, returns whether the IP is in the claimed bot operator's published IP ranges
- AI bot registry (JSON) — User-Agents and verification methods for AI companies' crawlers and fetchers, with a change log (CC BY 4.0)
03Planned
- Change alerts (email) — notifications of changes in standards or your site, without rewriting results
- Import of your own server logs, dedicated support — for the improvement verification pilot
- Checks of real AI assistants' interactions and the transaction stage — Phase 2
- Comparison with answers from AI apps — target apps to be announced at launch
Pricing
| Plan | Price (tax incl.) | Scans | Includes |
|---|---|---|---|
| Free scan | Free | One site, once | List of information AI can read, easy-to-follow fix list. No sign-up or card |
| Monitoring | ¥32,780 per site per month (¥29,800 excl. tax). Annual ¥327,800 (12 months for the price of 10) | Weekly | Per-entrance reports (Japanese PDF), fix list, detection of conflicts between standards, change alerts (planned). Same price for each additional site |
| Improvement verification (pilot) | Contact us | By consultation | Everything in Monitoring, before/after comparison of site fixes, import of your server logs (planned), dedicated support (planned) |
- Prices as shown on the official pricing page (October 2026). Displayed prices include 10% consumption tax
- Payment by invoice only; credit cards are not accepted. Annual plans are prepaid and renew automatically every year
- To cancel, email contact@migurei.ai before the next renewal date
How Parallel World differs from other LLMO/GEO tools
| Product | What it measures | Pricing guide | Provider |
|---|---|---|---|
| Parallel World | Whether AI crawlers, fetchers and agents can read your site (per published standards) | Free scan / monitoring ¥32,780/month (tax incl.) | Migurei (subsidiary of HBLAB Japan) |
| AIsearchmap | Your brand's share of mentions and cited sources in answers from ChatGPT, Gemini, etc. | Free plan / Enterprise from ¥50,000/month (custom quote) | CINC Inc. |
| LLMOinsight | Breaks exposure and recommendation rank in AI answers into a 100-point AI score, with improvement proposals | Free / from ¥9,800/month (excl. tax) | Letier Inc. |
| AI検索ドック | One-off diagnosis of how often you and competitors are named, from 30 questions × 10 answers per AI | ¥300,000 per diagnosis (excl. tax) | Stellagent Inc. |
- To learn how AI answers describe your company, answer-monitoring tools such as AIsearchmap or LLMOinsight are a better fit
- Parallel World measures the step before that: whether AI can read your site at all. Use it to find out whether the reason you don't appear in answers lies in robots.txt, structured data or prices rendered by JavaScript
- If you want the fixes done for you, Parallel World does not do that work; services such as genrank that also support improvements are an option
Pros and cons
Pros
- Records published standards and official tool results as they are rather than a proprietary score, so every result can be traced to its basis
- Shows conflicts and changes between standards with sources, and never rewrites past results
- The free scan needs no sign-up or card — just enter a domain name
- Read-only, with published limits on request count and spacing
Cons
- Measuring AI answers themselves (whether you are recommended) is planned; for now the focus is how readable your site is
- Full checks of the browsing/interaction and transaction stages (D and E) are Phase 2
- Applications by email only (form in preparation); payment by invoice only
- No fix work is done for you, so you or your web agency must make the changes
Operating company
According to the company page on the official website, the parent company HBLAB Japan Inc. was founded on November 28, 2017, has capital of ¥30 million, is located in Shibadaimon, Minato, Tokyo, and is represented by Nguyen Phi Tan. The website lists HBLAB group figures — HBLAB JSC in Vietnam (founded 2015, 650 employees), ISO/IEC 27001 and CMMI Level 3, and HBLAB Japan's PrivacyMark (obtained August 2024) — noting that these are published by the group, not by Migurei.
FAQ
Q. Can I use Parallel World for free? A. The free scan is available now: one site, once, with no sign-up or card. Weekly monitoring costs ¥32,780 per site per month (tax incl.).
Q. Does it support Japanese? A. It is built for the Japanese market; the website and the PDF reports are in Japanese.
Q. Will scanning affect or load my site? A. It is read-only and never submits forms, logs in or places orders. It follows robots.txt and sends at most 50 requests of its own per scan.
Q. Which AI does it measure? A. The bots published by each AI company (training bots such as GPTBot, ClaudeBot and Google-Extended; search bots such as Googlebot, Bingbot and OAI-SearchBot; answer-time fetchers such as ChatGPT-User and Claude-User).
Q. Can I ask them to implement fixes? A. No. The website states that Parallel World is a measurement service and does not carry out fixes or provide consulting.
Q. How do I contact support? A. Via contact@migurei.ai, with replies within two business days. Questions about the scanner's behavior are also accepted at research@migurei.ai.
This page is based on information published on Parallel World's official website (home, pricing, documentation, company page and llms.txt) as of October 2026. Please check the official website for the latest specifications and pricing.
