Privacy Policy

Last updated 3 August 2026

This policy covers the GuideGen browser extension and the website at guidegen.backpocket.website (and guide-gen.vercel.app, which serves the same site). It explains what we hold, why, and how to get rid of it.

1. The extension: local by default

When you record a guide, the extension saves the following to your browser's own extension storage, on your device:

• Screenshots of the pages you were on while recording
• The step description generated for each action, which you can edit
• The URL and page title of each step
• The position and size of the element you clicked, used to draw the highlight
• Your guide titles and step order

While a guide is unpublished, none of this leaves your machine. The extension makes no analytics, telemetry or tracking requests at any point.

The same list applies to catch-up capture, which records the same things on a site you have switched it on for and holds them for up to seven days so you can turn recent activity into a guide after the fact. It is off until you enable it, per site, and none of it is ever uploaded. Section 9a covers it in full.

From version 1.1 the editor is a page on this website rather than a page inside the extension, so that there is one editor to keep good instead of two. When you open it, that page asks the extension directly — browser to browser, over Chrome's externally_connectable channel — for the guides stored on your machine, one screenshot at a time as you scroll. Those screenshots are read into the page in your browser; they are not sent to our servers. The extension answers these requests only from this one website address, and from no other page.

Exports — HTML, Markdown, PDF, PowerPoint and video — are generated inside your browser and saved straight to your downloads folder. We never see them. Video narration is synthesized on your own machine by a speech engine bundled in the extension, so the text of your internal processes is never sent to a text-to-speech service.

2. What changes when you publish a guide

Publishing is an explicit action that creates a shareable link. Only then do we receive:

• The guide's screenshots, uploaded to our image host so a recipient's browser can load them
• The guide's text and structure — step descriptions, ordering, titles, and the page URLs and titles recorded with each step
• Your account identifier, so the guide belongs to you and only you can change it

Unpublished guides are never uploaded. Publishing one guide does not upload the others.

One consequence of sharing a link. When a published link is pasted into a chat app or a social network, that platform fetches the page to build a preview, and the preview shows the guide's title — to everyone in that conversation, whether or not they open it. The image beside it is a fixed GuideGen banner; a screenshot from your guide is never used as a preview image, so nothing from inside the guide is copied into a third party's preview cache. Shared guide pages are also marked not to be indexed by search engines.

3. Account data

From version 1.1 the extension asks you to sign in before it records. That does two things: it identifies whose guides the editor should show you, and it ties a published guide to you so that you — and nobody else — can update or delete it later. Signing in does not upload anything. A guide you have never published stays in your browser's own extension storage, on that machine.

We store your email address, your name, an internal user ID, and the timestamps of account creation and sign-in. We do not ask for a phone number, a company, or payment details. We do not use your email address for marketing.

You can sign in two ways. With email and password, the password is handled by Google Firebase Authentication and is never visible to us. With Continue with Google, we never see a password at all — Google confirms who you are and tells us your email address and name, nothing else. We ask Google for three things (openid, email, profile) and no access to Gmail, Drive, contacts or anything else in your Google account.

One consequence worth stating plainly, because it is a design decision rather than an oversight: unpublished guides cannot follow you to another computer. They are not on our servers, so there is nothing for us to sync. Published guides are the ones that travel.

Separately, if you export a guide that someone else shared with you, your email address is recorded against that guide for its owner to see — see section 4a.

4. Who processes your data

We use two service providers. We do not sell, rent or share your data with anyone else.

If you are in a region with data-transfer rules such as the UK, EU or India, note that publishing a guide involves transferring that guide outside your region. If that is a problem for your organisation, use the extension without publishing — every export format works entirely offline.

4a. If you export a guide someone shared with you

A guide's owner can switch on exports for that guide. If they have, and you sign in to export it, two things happen and you are told both before you sign in:

• The file is built on your own computer. The PDF, web page, Markdown, slide deck or video is generated in your browser from the images already on the page. We never receive it, and we never generate it for you.
• The owner is told that you exported it. They see your email address, which format you chose, and when. That is the whole record — not what you did before or after, not how many times you read it, nothing else.

If you would rather not be recorded, don't sign in: printing the page from your browser and saving the images work exactly the same either way, and leave no record with us.

That record belongs to the guide's owner. They can clear it, and it is deleted when they delete the guide. If you want your entry removed, ask them, or email us.

5. Public links

A published guide gets a link containing a long random identifier that cannot practically be guessed. Anyone holding that link can view the guide without signing in — that is the point of it — so treat it as you would any shared document link. Guide pages carry a noindex instruction so search engines do not list them.

You can unpublish a guide at any time, which immediately stops the link from working.

6. No tracking

This website sets no advertising or analytics cookies and runs no third-party trackers. If you sign in, your browser stores a session locally so you stay signed in. That is all.

7. Retention and deletion

• Unpublish a guide and its link stops working immediately.
• Delete a guide and its record is removed from our database immediately.
• Delete a guide's images — deleting a guide also deletes its images from our image host and requests removal from the content delivery network. CDN copies can take up to an hour to clear, and a browser that already downloaded an image may keep its own cached copy.
• Re-publish a guide and the superseded images are deleted from our image host as part of the update, with CDN removal requested. This matters if you re-publish specifically in order to redact something: the old image stops being retrievable. If that deletion fails, the images are still tracked against the guide and are removed when you delete it.
• Delete your account — email us and we will remove your account, all your published guides and their images within 30 days.
• Local guides are deleted from the extension itself, or by uninstalling it, which discards its storage. No copy exists elsewhere.
• Catch-up captures delete themselves after seven days, whether or not you have used them, and are capped in size before that. You can discard one at any point from your dashboard or the extension, and switching catch-up capture off for a site stops it holding anything new.

8. Your rights

Depending on where you live, you may have the right to access, correct, export or delete your data, or to object to our handling of it. Email us and we will help. You may also complain to your local data-protection authority.

9. Permissions the extension requests

The extension asks for access to all websites because it cannot know in advance which application you will want to document — it may be any internal dashboard or tool. The recorder does nothing unless you have explicitly started a recording, or have switched on catch-up capture for that specific site (see below), and it reads only the label and on-screen position of the element you click, plus the request summary described in section 9b.

It also asks for webRequest, which lets it see the network requests a page makes while it is capturing. This is observational only: it cannot block, redirect or modify a request. Section 9b covers what is recorded, and what is masked before it is stored.

9a. Catch-up capture

What it is. "Capture last 2 minutes" lets you make a guide out of something you have already finished. That is only possible if the extension was capturing while you worked, so on any site you switch it on for, GuideGen records each action and its screenshot as you go and discards them again — keeping only a rolling window of recent activity.

It is off until you turn it on, per site. There is no global switch in the extension's interface and it is never on by default. You enable it for one website at a time, and you can switch it off again at any point. While it is on, that site's pages show a small marker in the corner for as long as it is running.

What is kept, and for how long. The same information a recording captures — the label and position of the control, the page URL and title, and a screenshot of the visible page — held in the extension's own local storage on your device. It is capped, and anything older than seven days is deleted automatically whether or not you have used it. You can delete it yourself at any time from your dashboard or the extension.

It is never uploaded. Nothing held by catch-up capture is sent to us or to anyone else. It is not part of publishing: turning a capture into a guide creates an ordinary guide on your device, and that guide is only ever uploaded if you then press Publish on it, exactly as with a recording you started yourself.

Two things it will not capture. It does not run in private or incognito windows at all. And when a password field is focused, it keeps the written step but takes no screenshot — typed values are never recorded in either case, but a login screen is the one picture least worth keeping.

It also asks to create an offscreen document. That is a Chrome mechanism for doing work that needs a page but has no need to be seen: it is used to render the narrated video, because that requires a canvas, an audio engine and a video recorder. It has no access to any website you visit.

9b. The API log

What it is. While a recording is running — or on a site you have armed for catch-up capture — GuideGen notes the network requests the page makes, and attaches them to the step that triggered them. For each request that means the method, the address with its query values masked, the status code and how long it took. This is what turns "I clicked Save and it didn't work" into "clicking Save sent POST /api/orders and got a 500", which is the difference between a report someone can act on and one they cannot.

What it never records. The value of any header that looks like a credential, in any circumstance and on either setting below. Where a request carried an authorization, cookie, x-api-key or similarly named header, the name is kept and the value is replaced with a marker before it is stored — so a captured request shows that it was authenticated without carrying the token that authenticated it. There is no setting that turns that off. The same applies to values inside the query string and to obvious secrets in a sent body: a parameter called token keeps its name and loses its value, and a password field in a submitted form is replaced the same way. Your browser cookies are not readable by this mechanism at all.

The full exchange is a separate opt-in, and it is off. By default only the summary above is recorded. You can switch on "Include full API requests & responses" — once per recording, or per armed site for catch-up capture — and then GuideGen also keeps, for each request, its headers, the body that was sent and the body that came back, whether the request succeeded or failed. That is what lets it show you any request as a curl command with its response.

That command is a record of a call, not a way to repeat it: with the credential values masked and cookies absent, pasting it into a terminal will not sign in as you.

Everything else GuideGen records, you can see: a screenshot is a picture of what was on your screen, and you can black out parts of it before sharing. A response body is data you never looked at, which is exactly why this one is a decision you make each time rather than a default.

You can remove it. Each step's API log can be deleted from the editor, on its own, without affecting the rest of the guide. Turning the catch-up setting off also discards anything already held — the requests as well as the responses — not merely future ones.

Where it goes. Nowhere, unless you send it. The log lives with the guide on your device. Of the ways to get a guide out of GuideGen, only the AI handoff — the text you copy to paste into an assistant — includes it. HTML, Markdown, PDF, PowerPoint, narrated video and published shareable links all leave it out entirely, so publishing a guide never publishes its API log. When you paste a handoff into an assistant, the log goes wherever that conversation goes; that is the point of it, and it is worth reading what you are pasting.

The permission behind it. This is why the extension asks for webRequest. It is observational only — GuideGen watches requests, and cannot block, redirect or alter any of them.

10. Children

GuideGen is a workplace documentation tool and is not directed at anyone under 16.

11. Changes

If we add anything that changes what we collect, this page will be updated before that feature ships, and material changes will be announced to registered users by email.

12. Contact

GuideGen is operated by Gaurav Singh.
Email heygauravsingh@gmail.com for anything in this policy, including data deletion requests.