TL;DR
In 2026, 134 AI-in-education bills were introduced across 31 states. They disagree about almost everything except one principle: student work and student data shouldn't be shipped off to third parties by default. Meanwhile Google put a Gemini tab inside the Google Classroom mobile apps (June 2026), and the amended COPPA Rule has been in full effect since April 22, 2026. Heading into the 2026–27 school year, the least dramatic way to stay on the right side of all this is architectural: use tools that never send your work anywhere. Convert: Anything to PDF turns photos of your handwritten notes, lab CSVs, and Markdown notes into PDFs entirely on your own device — no upload, no account, no watermark, no size limit. One thing to know before you start, because you'll assume otherwise: we have no OCR. A photo of your handwriting becomes a picture inside a PDF, and that text is not searchable.
We are not lawyers. This is not legal advice. "Runs locally" is an architecture, not a compliance certificate.
What's actually happening in 2026
Three things landed, and they point the same direction.
134 bills, 31 states
134 AI-in-education bills were introduced across 31 states in 2026. They're not a coordinated program — they're 31 legislatures independently deciding that AI in classrooms needs rules. A few named examples:
- California AB 1159 — bars student data from being used for AI training.
- South Carolina H.B. 5253 — requires parental opt-in, and says AI can't replace licensed teachers.
- New York A.9190 — bans most classroom AI below 9th grade.
Read them together and the disagreements are loud: opt-in vs. opt-out, grade cutoffs, what counts as "AI." But underneath, they converge on a single instinct — a student's work and a student's data are not free inputs for someone else's system.
A Gemini tab in Google Classroom
In June 2026, Google shipped a Gemini tab in the Google Classroom Android and iOS apps — quiz generation, project brainstorming, tackling misconceptions. Whatever you think of that, note the timing: legislatures are drafting rules about classroom AI while classroom AI is arriving in the app on the phone in your pocket. The tools are moving faster than the statutes. That's not a scandal; it's just the situation you're studying in.
The amended COPPA Rule
Since April 22, 2026, the amended COPPA Rule has been in full effect. Three changes worth knowing:
- "Personal information" now expressly includes biometric identifiers.
- Separate verifiable parental consent is required for third-party disclosure — consent to use a service is not consent to hand data onward.
- A written data retention policy with a stated time period is mandatory. Not "we keep data as long as necessary." An actual period, in writing.
That middle one is the tell. The rule now treats disclosure to a third party as its own event requiring its own consent. That's the same instinct the state bills keep circling. (We wrote separately about the April 2026 COPPA deadline and saving compliance docs as PDF if you're on the school-admin side of this.)
The principle, and the honest version of our claim
Here's the thread: the direction of travel is that student work shouldn't leave by default.
A converter that runs entirely on your device is on the right side of that by construction, not by policy promise. There's a real difference between those two things:
- A policy promise says "we won't misuse your data." It's a sentence. It can be revised, reinterpreted, or rendered moot by an acquisition.
- An architecture says the file was never transmitted. There's no server copy to misuse, subpoena, breach, or repurpose — because there's no server copy.
That's a genuinely stronger position, and it's why we built it that way.
Now the part we won't blur: that is not a legal claim. We are not lawyers. Nothing here is legal advice. "Runs locally" is not a compliance certification. No state bill says "local tools are approved." Districts maintain their own approved-tools lists, and those lists are what govern you — not our architecture, and definitely not our blog.
So: check your district's approved-tools list. If a tool isn't on it, the fact that it's privacy-respecting doesn't put it on it. Ask. That's the whole advice.
What we'll say plainly is the architectural fact: when you convert a file with Convert: Anything to PDF, there are zero network requests during the conversion. The file is read by your browser, rendered by jsPDF on your machine, and saved to your disk. It doesn't touch us because there's no "us" in the path.
| Local conversion (on-device) | Online converter (upload) | |
|---|---|---|
| Where your file goes | Nowhere — stays on your disk | Uploaded to a server |
| Server-side copy exists? | No | Yes, at least temporarily |
| Third-party disclosure | None by construction | Depends on their policy |
| Works offline | Yes | No |
| Account required | No | Usually |
| Watermark on output | None | Often, on free tiers |
| File size cap | None | Usually, on free tiers |
Smallpdf, iLovePDF, and Adobe's online tools are all competent at converting files. The trade is the first row of that table. For a chemistry lab report that's a fine trade to think about; for a page of a classmate's data or anything with a name and a student ID on it, it's worth thinking about harder.
The thing you must know first: we have no OCR
Say it early, because students assume the opposite.
We have no OCR. None. When you photograph your handwritten notes and convert the JPG to PDF, what you get is a PDF with a picture of your handwriting inside it. The result is:
- Not searchable.
Ctrl+Ffinds nothing. - Not selectable. You can't copy a sentence out.
- Not readable by a screen reader.
- Not editable text.
It is a clean, ordered, shareable, printable PDF of your notes. That's genuinely useful — it's what a stack of loose paper wants to become. But it is a picture in a PDF, not typed text, and if you needed searchable text out of your handwriting, we are the wrong tool and you should stop reading and go find a real OCR tool. CamScanner and similar scanner apps do OCR; that's their job. (Weigh their privacy model yourself — a scanner app that OCRs in the cloud is doing exactly the thing the state bills are nervous about.)
The rule of thumb: text in gives you text out. Pictures in give you pictures out. Markdown, CSV, TXT, JSON, XML — those become real, selectable, searchable text in the PDF. JPG, PNG, WebP, BMP, SVG — those stay pictures.
Student workflows that actually work
Concrete stuff. All of it free, all on-device, no account.
Photos of handwritten notes and whiteboards → one PDF per topic
The end-of-semester reality: 60 photos of notes and whiteboards scattered through your camera roll in the order you happened to take them.
- Dump the photos for one topic into a folder. Rename them so they sort in the order you want —
01-,02-,03-. This is the whole trick and it takes two minutes. - Open Convert: Anything to PDF, add all of them, check the order in the list.
- Convert. You get one PDF per topic —
thermo-week3.pdf— instead of a camera roll.
JPG, PNG, WebP, and BMP all work. Whiteboard photos work well; shoot straight-on, because we don't do perspective correction either. There's a fuller walkthrough in turning phone photos of notes and textbook pages into a PDF.
Remember the OCR line: this PDF is pictures of your notes, ordered and merged. Perfect for reading and printing. Useless for searching.
Lab data CSV → a formatted table PDF
Your spreadsheet exports a CSV. The CSV looks like a wall of commas. Your lab report needs a table.
Export the data as CSV, drop it in, convert. It comes out as an auto-formatted table, and if you have 6 or more columns it automatically switches to landscape — because a 9-column table on portrait A4 is a ransom note. That table is real text: selectable, searchable, copyable.
One catch, stated plainly: Excel .xlsx is not supported directly. Export to CSV first. In Excel or Sheets that's File → Download / Save As → CSV. It's one extra step and we'd rather tell you than have you find out mid-conversion.
Paper sizes: A4, US Letter, US Legal, Tabloid, plus orientation. Whatever your program's submission guidelines say, one of those will match.
Markdown notes from Obsidian → a typeset PDF for submission
If you take notes in Obsidian, Notion, or any Markdown editor, you already have clean structured text. Point the converter at the .md file and you get a typeset PDF with real text — headings as headings, lists as lists, and content a professor's PDF reader can actually search.
No pandoc. No LaTeX install. No terminal. No "just install these four dependencies first." The Markdown to PDF guide for students goes deeper on the workflow.
Merging mixed files into one submission PDF
The one that saves the most pain. A typical lab submission is: 3 photos of the apparatus, 1 CSV of the readings, 1 Markdown writeup. Five files, three formats, one professor who wants one PDF.
Add all five. Arrange them in the order you want. Convert. They merge into one PDF in listed order — mixed formats, no limit on how many. The Markdown and CSV pages come out as real text; the photos come out as photos. One file, submitted.
There's no cap, no watermark, and no size limit, so a 40-page merged portfolio works the same as a 3-page one. Merging images, CSV, and Markdown into one PDF covers the ordering mechanics in detail.
Formats, at a glance
| Format | Supported? | Comes out as |
|---|---|---|
| JPG / JPEG, PNG, WebP, BMP | Yes | Picture in a PDF |
| SVG | Yes | Picture in a PDF |
| GIF | Yes — first frame only | Picture in a PDF |
| Markdown, TXT | Yes | Real, selectable text |
| CSV | Yes | Formatted table, real text |
| JSON, XML | Yes | Real, selectable text |
| HTML (local file) | Yes | Real, selectable text |
| Active web page | Yes | Real text, clickable links |
Excel .xlsx | No | Export to CSV first |
| Password-protected files | No | We can't read encrypted sources |
GIF is first-frame only. If a professor sends an animated GIF of a simulation, you get frame one. That's the honest limit, not a bug we're hiding.
When the sibling tool is the right answer
Here's a case we get wrong if we oversell: saving a page that's behind a login.
Your Canvas assignment page, your Google Classroom stream, a syllabus that only loads after you sign in — those aren't files on your disk. They're web pages that exist only inside your authenticated session. Convert: Anything to PDF can convert the active web page, but the tool built specifically for the login-protected, scroll-heavy, image-lazy-loading case is our other extension: Convert: Web to PDF. It uses your already-authenticated session, produces real text and clickable links, and has the extras that pages need — Article Mode, Load All Images, Remove Elements, Single Page Mode.
That's the honest split: files on your disk → Convert: Anything to PDF. Pages behind a login → Convert: Web to PDF. We wrote up saving webpages behind logins as PDF for exactly this. It's also the reason we ship four small extensions instead of one bloated suite — software should be small, and each tool should do one thing.
Worth doing before you need it: a course portal can go read-only after the term ends, and the syllabus you assumed would always be there is gone. Save it while you have access. Same logic as student PDF tools for finals week — the time to archive is before the deadline, not during.
What "free" means here
Because you've been burned before: no watermarks. No file size limits. No page caps. No account. No trial. No email address. Not freemium, not free-with-an-account, not free-until-you-hit-the-fifth-conversion-today. Just free.
That's not generosity, it's the manifesto: free is a feature, and privacy is the default rather than the upsell. And if we ever shut down, we publish the source and walk — we will not be acquired into adware.
Frequently asked questions
Will my handwritten notes be searchable after converting to PDF?
No. We have no OCR. A photo of handwriting becomes an image inside the PDF. You get a clean, ordered, printable document, but Ctrl+F will find nothing and you can't select the text. If you need searchable text from handwriting, use a dedicated OCR tool — CamScanner and similar apps do that, though many of them do it in the cloud, which is worth weighing against everything else in this post.
Does my schoolwork get uploaded anywhere?
No. File conversion happens entirely on your device via jsPDF, with zero network requests during conversion. The file is read by your browser and written back to your disk. There is no server copy because there is no server in the path. That's an architectural fact, not a promise — but it is not a compliance certification, and it doesn't override your district's approved-tools list.
Is this allowed under my state's AI-in-education law?
We genuinely can't tell you, and anyone who says otherwise is guessing. We are not lawyers and this is not legal advice. What we can say is that a converter is not an AI tool — it doesn't train on your work, doesn't send it anywhere, and doesn't generate anything. What governs you in practice is your district's approved-tools list. Ask your school.
Can I convert an Excel file for my lab report?
Not directly. Export to CSV first (File → Download → CSV in Sheets, or Save As → CSV in Excel), then convert the CSV. You'll get an auto-formatted table, and at 6+ columns it flips to landscape automatically so a wide dataset stays readable.
How do I control the page order when merging?
Files merge in listed order, so arrange them before converting. The easiest reliable trick for photos: rename them with a numeric prefix (01-, 02-) before adding, so they sort correctly from the start. There's no limit on how many files you can merge.
What about animated GIFs?
First frame only. If your source is an animated GIF, the PDF gets frame one and nothing else. PDFs don't animate, and we'd rather state the limit than let you discover it after submitting.
Can it convert a password-protected file?
No. We cannot read password-protected or encrypted source files. Remove the protection first using whatever tool applied it, then convert.
My professor wants one PDF but I have images, a CSV, and notes. Really one file?
Yes — that's the merge feature working as intended. Add all of them, order them, convert once. Mixed formats merge into a single PDF: the Markdown and CSV pages come out as real searchable text, the images come out as images. One file, in the order you chose, no watermark.
Bottom line
134 AI bills across 31 states, a Gemini tab in Google Classroom, and an amended COPPA Rule in full effect since April 22, 2026 all point at the same principle: student work and student data shouldn't be shipped to third parties by default. A converter that runs entirely on your device satisfies that by construction — there's no server copy to disclose, because there's no server. That's an architectural claim, not a legal one. We are not lawyers, "runs locally" is not a compliance certification, and your district's approved-tools list is what actually governs you — go check it.
For everything else: Convert: Anything to PDF merges your notes photos, lab CSVs, and Markdown into one submission PDF, free, with no watermark, no size limit, and no account. No OCR — a photo of handwriting stays a picture. Excel needs a CSV export first. GIFs give you frame one. Pages behind a login are Convert: Web to PDF's job.
That's the whole pitch. Small tool, honest limits, nothing leaves your laptop. Good luck this term.