The questions buyers raise first
Why not just ask Claude or ChatGPT whether I may scrape a site?
An assistant's answer is not evidence. It shows what a model said and not what the website said. The answer is not repeatable or versioned, so you cannot say which rules produced it. It is not sealed with the exact wording and the time it was read. An assistant does not fetch and check robots.txt, the reservation file, the headers and every policy page, and it will not tell you which of them it skipped. It may guess. It leaves no record that shows good faith later.
RightSignal reads the sources and quotes them word for word. Its own code checks each quotation against the page. It applies a signed and versioned algorithm, and it seals a receipt that anyone can verify.
Isn't public data free to use?
No. A page being public means you can reach it. It does not mean you may do anything you like with it. Copyright continues in freely accessible works. The Delhi High Court took that view in its interim ruling in ANI Media v OpenAI, which is under appeal. Section 43 of the Information Technology Act turns on the owner's permission and not on whether the page was open. EU law lets a rightsholder reserve public content from text and data mining. India's data protection law still covers much personal data that appears on public pages. Public is where the question starts.
Doesn't ANI Media v OpenAI mean training is fine in India?
It does not go that far. The ruling of 24 July 2026 refused an interim injunction. It took a first view that training on ANI's freely available content was fair dealing, and the court said its observations would not bind the final decision. The ruling is under appeal, with the appeal listed for 8 December 2026. It is a copyright ruling and did not rule on Section 43 or on a site's terms of use. It also rested on facts, which were that the content was freely available, that no paywall was passed and that the publisher had not opted out. Where a site has opted out, those facts are missing.
We only use Common Crawl. Why would we need this?
Using Common Crawl means someone else did the crawling. The question of what each site said remains yours. Common Crawl's crawler follows robots.txt, but the archive keeps no copy of a site's terms, and your use is training, which a site's terms may refuse. Sites also change after a crawl. Providers that serve the EU are already asked what they did about datasets from third parties. Trace reads what each site said at crawl time and what it says today.
Isn't a robots.txt check enough?
Robots.txt is one signal among several. It tells a crawler which paths to fetch and says little about use. A refusal of AI training, a licence requirement or a reference to Section 43 sits in a site's terms, which a robots.txt check never opens.
Will this slow my pipeline?
It need not. Most of what a site says applies to the whole site, so your crawler asks once per site and caches the answer until it expires. It does not call RightSignal before every page. The first read of a site takes as long as a polite read takes, so run your site list before the crawl begins. A check costs far less time than removing a source after training.
Who is liable if I collect after a decision?
You operate your own crawler, and you decide what to collect and how to use it. RightSignal provides evidence and decisions under the RightSignal Algorithm. It does not take on your responsibility for your collection, and the Terms of service set out what RightSignal is and is not responsible for. What RightSignal changes is what you are able to show. A sealed record that you read a site's signals and followed them is evidence of how you behaved.
Is this legal advice?
No. RightSignal is not a law firm and does not give legal advice or a legal opinion. Nothing it produces says that a collection or use is permitted in law. It gives you evidence and a decision under a cautious reading of the law, and your own lawyer advises you.
The product
What does RightSignal read on a site?
It reads robots.txt, text and data mining reservations, Content-Signal and Content-Usage directives, ai.txt, llms.txt and llms-full.txt, robots meta tags and headers, the site's terms, policies, copyright and licence pages, and any access barrier.
What are the five outcomes?
They are Collect, Collect on conditions, Hold for review, Information needed and Do not collect. The first two form the collect group, the next two form the hold group and the last is the stop group. Every outcome is named with its profile.
Does a Collect mean the site has given permission?
No. A Collect means RightSignal read the site's signals under your profile and found nothing that refuses or conditions your collection for that purpose. It is not a licence from the site's owner.
What happens when a site cannot be read?
RightSignal says so. Whatever could not be read is named as an evidence gap, and the outcome becomes a hold or a stop. A failed read is never treated as silence or as permission.
Which profiles can I choose?
You can choose India, EU or the India and EU Bridge, which applies the stricter of the two at every step. A strictest-caution option exists for the India profile and the Bridge.
Which purposes can I check today?
The RightSignal Algorithm decides search indexing, AI training, AI answer use, text and data mining and dataset building. The ready-made policies in the hosted service cover AI training under each profile and text and data mining under the EU profile. Policies for the other purposes are being added.
Which languages does RightSignal read?
It reads policy pages in English, Hindi, French, German, Italian, Spanish and Dutch. A policy page in another language is held for review.
Does RightSignal collect the content for me?
No. RightSignal reads a site's signals and policy pages so that it can decide. You run your own crawler and collect your own content.
Does RightSignal use AI to make decisions?
No. In live mode an AI reader can help to read a site's policy pages, and an assistant can help to find them. Code checks every quotation against the page, and the signed Algorithm reaches every outcome. Each decision says whether the AI reader took part.
Do you have a plugin or a software kit for my scraping framework?
Not yet. The integration today is the API, which any language can call. RightSignal will describe a plugin or a kit on this site only when it exists.
What is the status of each product?
Gate is in early access, by invitation. Trace is available by request. Watch and Matters are coming soon.
Receipts and evidence
What does a receipt prove?
It proves that the record has not changed since it was sealed and which key sealed it. The read times, the quoted lines and the outcome are therefore the ones recorded on the day.
What does a receipt not prove?
It does not prove that a decision is correct, that what a site published was true or that the site had the right to say it.
Can I verify a receipt without RightSignal?
Yes. The standalone verifier runs on your own computer with no connection to RightSignal.
How long do you keep receipts?
RightSignal keeps each live receipt for seven years by default. You should also export and keep your own copies.
What happens to my receipts when the Algorithm changes?
Nothing. A receipt names the version that made it and keeps verifying against that version. RightSignal never rewrites a receipt.
Can I see the rules behind a decision?
RightSignal publishes what the Algorithm decides and how its outcomes work. It discloses the full rules of a specific version to your counsel, an auditor or a court under a non-disclosure agreement.
Can anyone turn a stop into a collect?
No. Nothing in the hosted service lets a user turn a site's refusal into a collect.
RightSignal's own crawler
How does RightSignal's crawler behave?
It identifies itself as RightSignal, reads robots.txt first and keeps to a site's rate limits. It never passes a login, a paywall or a bot challenge, never solves a CAPTCHA and never disguises itself. It stops where a site asks it to.
Does it ever read against robots.txt?
In one disclosed case. It reads a site's own terms and policy pages even where robots.txt closes them, because that is where the site states its wishes. It records each time it does so and reads nothing else against robots.txt.
I run a website. How do I control RightSignal's crawler?
The crawler and site-owner policy explains how to recognise the crawler, how to slow it down, how to keep it away and how to complain.
Does RightSignal train AI on what it reads?
No. RightSignal does not train AI models on the pages it reads, does not republish them and does not sell them.
Accounts, pricing and data
How do I start?
Ask for access. RightSignal opens an organisation for you with test mode switched on, so you can build before you pay.
What is test mode?
Test mode makes decisions with test seals and without the AI reader. It is for building and testing an integration, and its output is not evidence for real collection.
What is a check, and what is an AI read?
A check is one page decided for one purpose, crawler and profile, with a sealed receipt. An AI read is one policy document read by the AI reader during a live check.
Is my dataset sent to RightSignal?
No. Gate needs the address of a page. Trace needs addresses and dates and never the content of your dataset.
Is anything sent to an AI model?
In live mode the text of a site's public policy pages can be passed to an AI model that helps to find and read them. Your own data is not. The Subprocessors page names the provider.
Where is my data stored?
The Security, Privacy and Subprocessors pages set out where the service runs and where records are kept.
Does RightSignal assess personal data?
No. RightSignal decides about access to pages and use of content. It does not assess personal data, and you need a separate basis for collecting any.
What if RightSignal gets a site wrong?
Tell us. Sites change, a site can publish a signal wrongly and a reader can miss wording. RightSignal checks the capture, tells you what it finds and corrects the reading for future decisions through a new signed version where one is needed. A sealed receipt is never altered.