Questions and answers
What governments ask us before they start, answered plainly. AI comes first because it is usually the first question: where it is used, exactly what it is sent, and how to turn it off for your whole organization.
The short version of where your files go is on how we handle your documents. Something not answered here? Ask us.
AI and your documents
Do you use AI to fix our PDFs?
For fixing a PDF, AI from another company is used in one place only: writing a description of each picture, where the fixed copy has tags and picture descriptions are on for your account. The report on each copy says whether its pictures were described. Adding a text layer to a scanned document, setting the title and language, naming form fields from the words printed beside them, building the tags and checking the reading order all run on our own servers, and none of them sends your document anywhere.
The one other place is the web page (HTML) version of a document. Our own layout reader, which runs on our servers, converts it where your account has one. A document that reader cannot read, or one longer than it takes, is converted by AI instead, unless AI is turned off, in which case it is not converted. Where picture descriptions are on for that work, the pictures in a web page version may be described by AI too, whichever way the version was made.
Exactly what is sent to the AI, and when?
For a picture description: that one picture, the document's file name (for a web page version, the document's title instead: the one our layout reader read from it, or the one AI wrote for the version), the picture's caption if it has one, and any words already read inside the picture. Never the rest of the page. A picture found inside a scanned page is sent as that picture alone, with a note that it came from a scan, never the whole page. What comes back is a short description, and whether the picture looks like decoration. In a fixed PDF that is only a suggestion, and a person marks a picture as decoration in the review. In a web page version it is applied: a picture AI calls decoration is given no description there, so check the version before you publish it.
For a web page version AI writes: the document's text, page by page, and its file name. Its pictures go separately, as above, where picture descriptions are on.
For a worked example on a website finding: the public HTML of that one element of your website and the rule it failed. No document content.
Nothing is sent during a website scan, and nothing is sent unless somebody on your account asked for a fix, a conversion or a worked example.
Which AI company, and which model?
Claude, made by Anthropic, used through Anthropic's commercial API. No other AI company receives anything from your documents or your website.
Is our document used to train AI?
Under Anthropic's commercial terms, what we send through its API is not used to train its models. We do not use your documents to train anything either. How long Anthropic keeps what it receives is set by those same commercial terms. With AI turned off, Anthropic receives nothing from your documents at all.
Can we say no to AI?
Yes. Anyone signed in to your organization's account can turn AI off for the whole organization on the Account page, or ask us and we will do it. You can also choose "Don't use AI on these files" when you send one set of files, and leave AI on for everything else.
With AI off, nothing from your documents or your website is sent to any AI company, and no AI model writes descriptions or other words of its own into your files.
Does anything use machine learning when AI is off?
Yes. Two pieces of software that run on our own servers use machine learning: text recognition, which reads the words on a scanned page, and the layout reader, which works out where the headings, paragraphs and tables are when a document is converted to a web page. Neither sends your document anywhere, and neither writes words of its own: they read what is printed on the page.
If your policy counts that as AI too, tell us before you start and we will go through it with you.
If AI is off, is our PDF still fixed?
Yes. Everything else runs exactly as it does with AI on. Pictures come back without a description, and where the fixed copy has tags, each picture gets its own card in the review for a person on your team to describe.
Who writes the picture descriptions when AI is off?
You do, in the guided review, wherever the fixed copy has tags: each picture there has a card where you type a description, or mark the picture as decoration so a screen reader skips it. What you type is written into the PDF you download. Descriptions are plain text, and nothing you type is sent to any AI.
A copy without tags has nowhere to hold a description; the review says when a copy has none, and why.
Do we pay for AI work that was not done?
No. Picture descriptions are part of the credit a page already takes, so turning AI off does not change what a page costs, and a document that is not fixed or not converted costs nothing.
What about the web page version when AI is off?
Where your account converts documents with our own layout reader, it still does, on our servers. A document that reader cannot read, or one longer than it takes, is not sent to AI: it is reported as not converted, and it costs nothing. Where your account has no such reader, the web page version is not offered while AI is off. Fixing the PDF itself calls no AI model and still works.
When AI writes a description, how do we know?
In a fixed PDF, each description AI wrote is written into the copy you download, and labelled as written by AI beside its picture in the report and in the review. It keeps that label after a person accepts it.
The PDF itself does not mark a description as AI's, so read each one in the review before you publish. You can rewrite it or remove it there.
A web page version has no review. Where AI described its pictures, each description is written in as the picture's text alternative and is not labelled there, so read the version before you publish it.
How accurate are AI's picture descriptions?
A description says what is visible in the picture, not why the picture is on the page, so it can miss the point or be wrong. That is why, in a fixed PDF, each one is labelled as AI's in the report and the review, for a person to read before the document is published, and why a web page version should be read before it is published too.
Does AI decide the headings, reading order or form labels?
In the fixed PDF, no AI model does. Headings are proposed from the size and boldness of the type, form labels from the words printed beside each box, and the reading order is checked against where the text sits on the page. Each proposal is listed in the report with the evidence for it.
In a web page version, the headings come from our own layout reader where it converted the document, and from AI where AI wrote the version. Check them as you would any draft.
Is AI on for a new account?
Yes. Until somebody turns it off, AI may be used in three places where your account has them: a description of each picture in a fixed PDF that has tags, a web page version of a document our own reader cannot convert (and a description of each picture in a web page version, however it was made), and a worked example when somebody presses "Show an example fix". Turn AI off on the Account page before you send your first file, and nothing from your files or your website is ever sent to AI.
Our organization has its own AI policy. How does this fit?
Turn AI off, and no AI company receives anything from your documents: they are processed on our own servers and kept with the hosting and database providers our privacy page lists. Where a fixed copy has tags, each picture description is then written by a person on your team, as plain text, in the review.
Can we change our mind later?
Yes, at any time, on the Account page. Turning AI off applies from the next document: work already running finishes the documents it is on and sends nothing more to AI after them.
Documents already handed to AI to be converted later are stopped where they have not started, and nothing produced for them is collected or written into your documents. What was already sent cannot be recalled. Descriptions AI already wrote into a fixed copy stay in that copy until it is removed or fixed again, and you can rewrite or remove each one in the review.
Does anything else in the product use AI?
Only the optional "Show an example fix" button on a website scan finding. It sends the public HTML of one element of your website and the rule it failed, and no document content. With AI off that button is not offered; the guidance on each finding, which is the rule itself, still is.
Can we write descriptions with our own approved AI tool?
Yes. Whatever you type into a picture's card in the review is written into the PDF as plain text, wherever the words came from. If your organization uses an approved AI tool, you can paste its description there and check it as you would any draft.
Can we use our own AI key or provider?
Not today. AI work runs on our account with Anthropic, and you can turn it off entirely.
Can we see exactly what was sent to the AI for a document?
For a fixed PDF, the report lists every description AI wrote, beside the picture it was shown. A web page version's report gives only counts, so read its pictures' descriptions in the version itself. The request itself is not shown on screen; what goes into one is described in full above.
Your data, storage and security
Where are our documents stored, and for how long?
Every PDF you upload is held in its upload set for three days after you add it. A PDF you fix is also kept, beside its fixed copy, for a week after fixing finishes, so you can compare and download them. Once a window closes those files are no longer available, and they are removed from our database within about an hour. The record that a PDF was fixed, meaning its name, the date and what happened, stays, so you can show what was done.
Converting a PDF to a web page works on a copy of it. That copy is removed in the same step that saves the web page, and the PDF you uploaded stays in its upload set for the three days above. The web page version can be downloaded for three days after it is made, then is removed the same way.
Assessments, findings and document lists are kept for as long as you have an account, and deleted when you ask. Our hosting and database providers' infrastructure is in the United States.
Who can see our documents?
In the product, only people signed in to your organization's account. Another organization asking for one of your files is told it does not exist, and the support and billing screens our own staff use cannot open your documents.
Where AI is on, the parts listed in the AI questions above are sent to Anthropic to be processed.
Who at MissionLayer can see our documents?
In the product, nobody at MissionLayer can open your documents: the support and billing screens our staff use cannot reach them. Only the people who run the service can reach the database directly, and they open a customer's data only to fix a problem you have reported.
Our documents contain personal information (FERPA, health records, criminal justice). Can we use GovAccess?
GovAccess is built for documents your office publishes. Do not upload records that are confidential or that contain residents' personal information unless your office has approved it. We do not sign a HIPAA business associate agreement and do not meet CJIS requirements. Turning AI off keeps every document on our own servers and the providers listed on our privacy page.
Does the free PDF page keep our file?
Your file is never saved. The fixed copy waits in our server's memory for up to fifteen minutes until you download it, and is gone once you do. The free page calls no AI model; where your PDF has no text, text recognition reads it on our own servers.
If you give us your email address to download the copy, that address is kept as a contact, and we email you at most once to ask how it went. Our privacy page says how long it is kept and how to have it removed.
The free page is one of our public pages, so the advertising and visitor-measurement tags our privacy page lists run on it. The panel that shows your file's name and results is hidden from the session recording.
How do you sign people in?
With a six-digit code emailed to the person, which works only in the browser that asked for it and expires after ten minutes. No customer is given a password. A copy of our database holds no working sign-in credential: the session tokens we keep are stored only as one-way digests.
Is our data encrypted?
Everything between your browser and our service travels over HTTPS. Secrets you give us are encrypted with a key held outside the database, and session tokens are stored only as one-way digests.
Do you have SOC 2 or a VPAT?
No. We do not hold a SOC 2 report, and we have not published a VPAT. Our accessibility statement lists the known shortfalls in our own screens. We answer security questionnaires directly: send yours through the contact page.
What other companies handle our data?
Our hosting and database providers, Stripe for invoices, Resend for email, Anthropic only where AI is on, and Google and Microsoft for advertising and visitor measurement. The complete list, with exactly what each receives, is on our privacy page.
Advertising and visitor-measurement tags run only on our public pages, never on a screen that shows your documents, findings, invoices or support messages. Visitor counts and session recording also never run for someone who is signed in. The advertising tag is decided by the page alone, so it runs on a public page whether or not the visitor is signed in.
One thing reaches Google from outside those pages: when an account that came from one of our advertisements pays an invoice, we send Google that advertisement's click identifier, the amount and when it was paid, and nothing else.
The free PDF page is one of those public pages; on it, the panel that shows your file's name and results is hidden from the session recording.
Will you tell us before adding a new company that handles our data?
Yes. We update the list on our privacy page before a new company receives any of your data, and we email each account's contact about the change.
Will you sign our data processing agreement, and how do you notify us of a breach?
Send us your data processing agreement and we will review it. If we learn that your data was accessed without authorization, we tell your account's contact without undue delay, and within 72 hours, with what we know.
Can we export our data or have it deleted?
Fixed PDFs, reports and the ZIP of a set download from Your PDFs while they are kept. A scan's list of documents exports as a spreadsheet. Your organization's assessments, findings, document lists and support tickets have no delete button: write to us and we delete them.
Does the website scan use AI, or change anything on our site?
No AI model is called in a scan. It is passive: it follows public links with ordinary page requests, does not sign in, does not submit forms, obeys your robots directives and waits between requests.
Does fixing replace our original?
No. The fixed copy is a new file, and your original stays exactly as it is. Our copies are removed after the windows above, so keep the downloads you need; your own records schedule governs your own files.
What gets fixed, and what a person still checks
What does GovAccess actually fix in a PDF?
Where a PDF has no text at all, it adds a text layer by reading the scanned pages. It sets the title and language, names form fields from the labels printed beside them, writes tags where the copy can carry them, and checks the reading order of those tags.
A copy can come back without tags, for example where a scanned page had no text that could be read, for an XFA form, or where tags cannot be written for your account. The review says which, and why.
Will the fixed copy be fully accessible?
Not necessarily, and no automated tool can promise that. Every fixed copy comes with a list of what a person should still check before you publish it, and that list is never empty.
What is the guided review?
A short set of cards, one for each thing a person should decide: picture descriptions where the copy has tags, form labels nobody could read, the title, the language, and places where the reading order may be wrong. You answer them on screen and download a copy with your answers written in. It calls no AI model and costs no credits.
How do you check reading order?
We walk the tags in the order a screen reader would and compare that with where the text sits on the page. Standard checkers, veraPDF included, cannot see reading order at all, so we run our own check on every copy that has tags. When you fix PDFs on Your PDFs, we also run the standard veraPDF check and show the two side by side, because neither one replaces the other.
Does a passing check mean the document is fine?
No. A check that passes tells you what that check looked at, nothing more. A check that could not run is reported as not checked, never as passed.
Can you make a web page (HTML) version instead of a PDF?
Yes, where your account converts documents. A web page version can be pasted into your website or downloaded, and it is checked in a browser each time you open its report, or marked as not checked where that check could not run. Forms in it can be filled in and printed or saved, but they do not submit anywhere.
Who writes it, our own layout reader or AI, is explained in the AI questions above.
Which kinds of file can it fix?
PDFs only. Word, Excel and PowerPoint files found on a website are listed but not opened.
Which languages can it set or read?
The guided review can set a document's language to any of these: English, Spanish, Chinese (Simplified), Chinese (Traditional), Vietnamese, Korean, Tagalog, Russian, Arabic, French, Haitian Creole, Portuguese, Japanese, Somali, Hmong, Navajo, German, Ukrainian, Persian, Polish, Italian, Hindi, Punjabi, Khmer, Lao, Amharic. Text recognition, which reads scanned pages, reads English.
How many PDFs can you handle, for example 500?
A set on Your PDFs is fixed one document at a time on our servers. Tell us your volume and we will give you an estimate before you start. Credits are held when you press Fix, so the set is priced before it starts.
Do you offer a done-for-you remediation service?
No. GovAccess fixes what software can fix and tells you what a person should still check. Your team, or a contractor you choose, does the rest.
Forms, scans, signed and locked PDFs
Can it fix fillable forms?
It names each field from the words printed beside or above it, so a screen reader can announce "Last name" rather than "Text1". A field that already has a name is left alone, and a box with nothing printed near it is listed for a person to name in the review.
What about scanned documents?
Where a PDF has no text at all, we read the words on our own servers and add them as a text layer. A PDF that mixes typed and scanned pages is not read this way yet.
When you fix PDFs on Your PDFs or on the free page, pictures inside scanned pages are found and redrawn in place so the page looks the same. A scanned page with anything drawn over it, such as a redaction box, is left as it was rather than redrawn.
What happens with a signed or password-protected PDF?
We do not change it, because a change would break the signature or the protection. The screen says why, and it costs nothing. For a signed document you can make a web page version instead.
Accessibility rules and the law
What does ADA Title II require for our PDFs?
The Department of Justice's Title II rule adopts WCAG 2.1 Level AA for the web content of state and local governments, including the PDFs on your website, with exceptions for some archived and older documents.
When are the ADA Title II deadlines?
The Department of Justice's rule gives state and local governments with a population of 50,000 or more, other than special district governments, until April 26, 2027, and smaller governments and special districts until April 26, 2028. Check the current dates with your counsel: the rule and its dates can change.
Does using GovAccess mean our documents meet the ADA?
No. GovAccess helps you get ready, and no software can make that legal determination. It fixes what software can fix in each PDF and tells you, file by file, what a person should still check. Whether a document is required, and whether it meets the rule, is your organization's decision.
Do we have to fix every old PDF?
Not necessarily. The rule has exceptions, for example for archived content, and for some documents posted before the compliance date that nobody now needs to apply for or use a service.
Where your account lists your website's documents, it sorts them into likely in scope and likely excepted, with the reasons. That is a starting point for your own decision, not a legal one.
Do you support Section 508 or federal agencies?
No. GovAccess is built for state and local governments under ADA Title II, and is not offered to federal agencies.
Buying as a government
How do we pay?
You buy a pack of credits and get an invoice with a secure payment page, where a card, a purchasing card included, or a bank transfer both work. Invoices are due in 30 days unless we agree otherwise.
Can we pay against a purchase order?
Yes. Put your PO number and accounts-payable address on the invoice when you buy; the credits arrive when it is paid. Ask us for a quote or a W-9 through the contact page.
Do you have a W-9?
Yes. Ask for one through the contact page and we send it with a quote.
Are you on a cooperative contract, or can this be a sole-source purchase?
Not on a cooperative contract today. Ask us for a quote and a sole-source letter if your purchasing rules allow one.
Do credits expire? Is this a subscription?
Credits never expire, and nothing renews or charges automatically.
What does a page cost, and what if a PDF cannot be fixed?
The pricing page shows the credits a page takes and the pack prices. Credits are held when you press Fix, and only the pages actually fixed are charged; a PDF that cannot be fixed costs nothing.
Do you store our card?
Card and bank details never reach GovAccess. Stripe, which runs the payment page, may keep a card on file with Stripe for your next payment.
Do you refund unused credits? Do nonprofits or small towns get a discount?
Credits never expire, so we do not refund unused credits. Write to us about pricing for a small government.
What happens if we stop paying, or if MissionLayer closes?
Nothing is charged automatically, so there is nothing to stop. Your credits stay on your account. If we ever close the service, we tell every account at least 60 days ahead, and until then every file you can download stays downloadable for as long as it is normally kept.
Do you carry cyber insurance?
We do not carry cyber insurance today. If your purchasing requires it, tell us and we will arrange coverage.
Who is the legal entity, and what is the address?
MissionLayer GovAccess is provided by MissionLayer Systems LLC, 3267 S Tytus Ln, Saratoga Springs, UT 84045.
Can a school district, library or nonprofit use it?
Yes. Anyone with a working email address can open an account. Which accessibility rules apply to your organization is a question for your own counsel.
Getting started
Can we try it before buying anything?
Yes, two ways. The free PDF page fixes up to 5 pages of one PDF with no account, and calls no AI model. A free account comes with credits to start with and asks for no payment details.
How many files can we try on the free page?
A few an hour, and the page shows how many you have left. A longer document is cut to its first 5 pages, and the page says so.
Does the free page use AI?
It calls no AI model. Where your PDF has no text, text recognition reads the words on our own servers, and sends your file nowhere. Pictures it finds are left without a description and counted on the result.
Support
How do we get help?
Signed-in accounts raise a ticket under Support, and every message on it is visible to both sides. Anyone can also reach us through the contact page.
How fast do you respond?
We aim to answer a new question or problem within two business days, anything that stops you working within one, and a request for a new feature within three. Each ticket shows its target when it is raised. A feature request gets a first answer, but we do not promise delivery dates.