How to Use fileForensics AI for Beginners

Your AI assistant just read a PDF and gave you a confident answer. But what if that PDF was edited? What if a number was changed, a clause was swapped, or the whole document was written by another AI and passed off as a real report? Your AI does not know. It reads what the file says — it never asks whether the file can be trusted.

fileForensics AI fixes that blind spot. Launched on Product Hunt on November 4, 2026, it is a trust check for every file your app or AI agent acts on. As its makers put it: extraction tells you what a document says; fileForensics tells you whether to believe it. It looks for signs of editing, AI-generated content, tampering, and details that do not add up — then returns its findings with the evidence: what it found, where it found it, and why it matters. No black-box score. In this beginner guide, you will learn exactly how to use fileForensics AI, from your first file check in the developer portal to connecting it to Claude, Cursor, or Codex with MCP and running one-call checks with its API. No technical background needed.

What Is fileForensics AI?

fileForensics is a file-trust tool from the fileAI platform. It answers a question that most AI tools skip entirely: before your AI reads, quotes, summarises, or acts on a file, is that file what it claims to be?

Think about how AI assistants work today. You hand ChatGPT, Claude, or an AI agent a document, and it extracts the text and works with it. The extraction is usually accurate — but extraction only tells you what the file says. It says nothing about whether a figure was altered after the document was created, whether a paragraph was generated by AI and dressed up as a human-written report, or whether two pages of the same document contradict each other. Your AI treats a doctored invoice exactly the same as a genuine one, because it was never asked to check.

fileForensics sits in front of that step. You give it a file, and it examines the file for trust signals instead of just reading the words. Here is what it checks, in plain language:

• Signs of editing — traces left behind when text, numbers, or images in a file were changed after the file was created.
• AI-generated content — patterns that suggest part or all of the file was produced by an AI model rather than a person.
• Tampering — inconsistencies that point to deliberate manipulation, such as a figure that was replaced or a page that does not match the rest of the document.
• Details that do not add up — dates, totals, names, or claims inside the file that contradict each other or the file's own metadata.

The important part is how it reports back. Instead of a single mysterious trust percentage, fileForensics returns findings with the evidence behind each one: what it found, where in the file it found it, and why it matters. You see the reasoning, so you can judge for yourself instead of blindly trusting a score.

Step 1: Understand Why Your AI Needs a File Trust Check

Before you open the tool, get clear on the problem, because it decides when a file check is worth your time. AI agents are now doing real work with real files: reading invoices, screening job applications, summarising contracts, verifying research, and answering customers from uploaded documents. Every one of those jobs assumes the file is honest. That assumption is the risk.

Here are everyday situations where a file check earns its place:

• An invoice or receipt arrives by email. The total looks right, but was a figure changed before it was sent to you? A trust check catches the edit before you pay.
• A job applicant sends a CV or certificate. The document looks polished, but is it genuine, or was it assembled by an AI and formatted to look official?
• A contract or agreement lands in your inbox. One clause reads differently from the version you discussed. A check shows you exactly what changed and where.
• A research paper or report is quoted to you. The numbers are impressive, but do they add up inside the document itself, or does one page contradict another?
• A screenshot is offered as proof. Screenshots are easy to fake and hard to verify by eye. A file check looks at the pixels and metadata instead of taking the image at face value.

The rule of thumb is simple: if a decision, a payment, or a public claim will rest on a file, check the file first. If you are only chatting with AI about a file for your own notes, a check is optional. Start with the high-stakes files, and the habit will form on its own.

Step 2: Open the fileForensics Developer Portal

Your starting point is the fileForensics developer portal — the no-code way to run your first checks without installing anything or writing code.

Search for "fileForensics" and open the official page from the results, or find its launch post on Product Hunt from November 4, 2026, which links out to the product. The portal runs in your browser, so there is nothing to download and no app to install. Create your account with your email, or sign in with the option the portal offers.

Once you are inside, take two minutes to look around before uploading anything. Find where your checks will be listed, where the findings for each file appear, and where your API key lives if you plan to use the API later (Step 6). Knowing where those three things are now saves you from hunting for them when you are in a hurry later.

The portal is the right place to learn the tool. Everything it can do — the checks, the evidence, the findings — shows up here in its clearest form. The MCP and API routes you will meet in Steps 5 and 6 are the same checks, delivered to your AI tools instead of your browser.

Step 3: Upload Your First File and Run a Trust Check

Pick a file you already have. A PDF invoice, a Word document, a report, or an image works. For your first run, choose a file you trust — something you created yourself or received from a known source. That gives you a baseline: you will see what a clean result looks like before you ever test a suspicious one.

Here is the first-check routine:

• Upload the file in the portal, either by dragging it in or choosing it from your computer. Use the original file whenever you can, not a screenshot or a photo of it, because copies and re-saves strip away the traces the check looks for.
• Start the check and wait for it to finish. The tool examines the file's content, structure, and metadata, so give it a moment rather than refreshing the page.
• Open the findings when the check completes. You will see a list of findings rather than a single score. Read each one slowly on your first run — this is where you learn what the tool actually notices.
• Compare with what you know. Since this is a file you trust, most findings should be minor or empty. If something surprising appears, look at the evidence it cites before deciding whether it matters.

Run two or three familiar files before you test anything suspicious. The goal is to learn the tool's language: what a normal finding looks like, what a warning looks like, and how much detail the evidence gives you. Ten minutes of practice here makes every later check faster and more confident.

Step 4: Read the Findings — What, Where and Why

This is the skill that matters most, because a file check is only useful if you can read its answer. fileForensics deliberately avoids a black-box score — one number that tells you "87% trustworthy" and hides the reasoning. Instead, every finding comes with three parts: what was found, where in the file it was found, and why it matters.

Read a finding like a short report:

• What — the observation itself. For example, "a figure on page 2 shows signs of being replaced after the document was created," or "this paragraph matches the writing patterns of AI-generated text."
• Where — the exact location. A page number, a section, or a region of an image. This is what lets you open the file yourself and look at the same spot with your own eyes.
• Why — the reason it matters. A replaced total changes what you owe. An AI-generated paragraph changes how much you trust the author's claims. A mismatched date changes whether a document is still valid.

Treat findings as evidence to weigh, not as a verdict. A single minor finding on a file from a trusted source is usually noise — a compression artefact, a scanned page, a file that passed through three apps before reaching you. Several findings pointing at the same page, or one finding that lines up with something you already suspected, is a different story: open the file, look at the spot the finding names, and decide with your own judgement.

The practical habit is to ask three questions about every check: does anything here change the decision I was about to make, does anything here contradict what the sender told me, and is there anything here I should ask the sender about before I act? If the answer to all three is no, you can proceed. If any answer is yes, pause and verify.

Step 5: Connect fileForensics to Claude, Cursor or Codex with MCP

The developer portal is great for checking one file at a time. But the real power of fileForensics shows up when your AI agent checks files by itself, before it acts on them. That connection runs through MCP — the Model Context Protocol — and fileForensics ships a hosted MCP server for it.

MCP is simply the standard way AI tools like Claude, Cursor, and Codex connect to outside services. Once the fileForensics MCP server is connected, your agent gains a new ability: it can run a file trust check as part of its own work, without you opening the portal at all.

Here is the setup in plain steps:

• Open your AI tool's MCP or integrations settings. In Claude, Cursor, or Codex, look for the settings area where external servers or MCP servers are added.
• Add the fileForensics MCP server using the connection details shown in the developer portal. The portal provides the server address and any key you need — copy them exactly rather than typing them from memory.
• Save and restart the connection if your tool asks for it, then open a fresh chat with the agent.
• Test it with a real request. Upload or point the agent at a file and ask it, in plain words, to check the file before using it — for example: "Check this invoice for tampering before you summarise it." Watch what the agent reports back: it should return the findings, with the what, where, and why, instead of just a yes or no.

Once this works, change how you instruct your agents. Instead of saying "read this file and act on it," say "check this file first, then act on it only if the check is clean." That one sentence turns fileForensics from a tool you remember to use into a guardrail that is always on. If you run a small team and want that guardrail built into a wider workflow, an AI automation services partner can wire the check into your existing agent setup so every file passes through it automatically.

Step 6: Run One-Call File Checks with the API

If you build apps, run automations, or process many files at once, the API is the fastest route. fileForensics is designed around a simple idea: one call, one file, full findings back. You send the file, and the API returns the same evidence-backed findings you see in the portal.

The beginner's path looks like this:

• Get your API key from the developer portal. Treat it like a password: never paste it into a chat, a document, or a public post. If you use AI chat tools while working with keys, our guide to PasteGuard AI shows how to keep secrets like API keys out of your AI chats.
• Send one file in one call to the file-check endpoint, following the quick-start example in the portal's API docs. Start with a single small file so you can see the exact shape of the response.
• Read the response the same way you read the portal — findings with what, where, and why. Your code should pass the findings through to a human whenever anything is flagged, rather than silently acting on a file that failed its check.
• Decide your rule in advance. For example: clean files proceed automatically, flagged files wait for a human to review. Writing that rule down before you automate anything is what keeps the API a safety tool instead of a rubber stamp.

You do not need to be a developer to understand the value here: the API means the check can happen inside your own product or workflow, at the moment the file arrives, instead of depending on someone remembering to open a portal. Start with the portal, move to MCP when your agent needs it, and reach for the API when files arrive faster than a person can check them.

Step 7: Check PDFs, Images and Documents the Smart Way

Different files hide different problems, so the way you check them changes a little by file type. The tool does the analysis; your job is to give it the best possible file and to know what to look for in the results.

• PDFs — the most common target. Watch for replaced figures, hidden text layers, pages that were swapped in from another document, and totals that do not match the line items. Always upload the original PDF, not a print-to-PDF copy, because re-saving a PDF can erase the very traces the check looks for.
• Images and screenshots — watch for AI-generation signals, edited regions, and metadata that contradicts the story. A photo claimed to be taken yesterday should not carry metadata from last year. Remember that a screenshot of a document is a weaker starting point than the document itself, because the screenshot process strips detail.
• Word documents and text files — watch for mixed writing patterns, such as one section that reads human and another that reads machine-generated, and for metadata that does not match the claimed author or date.
• Scanned documents — expect more noise. Scanning adds compression and artefacts that can look like tampering to any checker. Treat findings on scans with extra caution and weigh them against the source: a scan from your own scanner is different from a scan forwarded by a stranger.

One habit covers all four types: keep the original file. When someone sends you a file that matters, save it exactly as it arrived before you forward it, convert it, or upload it anywhere. Every conversion is a chance to lose the evidence a trust check needs.

Step 8: Build a File-Trust Habit Into Your AI Workflow

A tool only helps if it is used at the right moment, and the right moment for a file check is before your AI acts on the file — not after. The check is a gate, not an autopsy. Build it into the places where files enter your work.

Start with three gates that cover most situations:

• Before your agent acts. Any file an AI agent will use to send an email, make a booking, process a payment, or update a record gets checked first. This is the MCP setup from Step 5, turned into a standing instruction.
• Before you quote or publish. Any file you will cite in an article, a report, or a social post gets checked before the words go out. A wrong number in a published post is much harder to fix than a flagged file in your inbox.
• Before you pay or sign. Any invoice, contract, or agreement gets checked before money moves or a signature lands. This is the highest-value gate, and it takes the least time.

If you run a business and files flow through several people, write the gates down as a one-page rule and share it with the team. The rule does not need to be complicated: who checks, which files get checked, and what happens when a file is flagged. A digital agency can help you turn that one-page rule into a documented workflow if file handling is a regular part of your client work.

The point is not to become suspicious of every file. It is to make the check automatic for the files that carry real consequences, so trust becomes a step in the process instead of a feeling you hope is right.

Step 9: Know the Limits Before You Trust the Verdict

fileForensics is a strong tool, but no file checker is perfect, and knowing its limits is part of using it well. The tool itself is honest about this: it returns findings with evidence, not a final verdict, precisely because file trust is a judgement call.

Keep these limits in mind:

• No checker catches everything. A carefully made fake can pass, and a genuine file can be flagged. Treat the findings as a signal to investigate, not as proof in either direction.
• Editing leaves traces, but so does ordinary handling. Compressing, converting, scanning, and re-saving a file can all create signals that look like tampering. Context matters: a file that passed through WhatsApp will look different from the same file sent as an email attachment.
• AI-generated text is hard to pin down. Heavily edited AI text can look human, and formal human writing can look machine-generated. A single "AI-generated" finding on its own is a reason to look closer, not a reason to accuse anyone.
• Evidence beats a score. This is why the no-black-box-score design matters. When a finding gives you the what, the where, and the why, you can open the file and verify it yourself. When a tool only gives you a number, you cannot.
• High-stakes decisions need more than one check. If a file decides a payment, a legal matter, or a public claim, pair the file check with source verification: contact the sender, ask for the original, or compare with an independent copy. The file check tells you where to look; the source check tells you what is true.

Beginners make two mistakes here. The first is treating a clean result as a guarantee — it is not, it is the absence of found problems. The second is treating one flagged finding as proof of fraud — it is not, it is a reason to look. Stay in the middle: read the evidence, verify the spot it names, and then decide.

Step 10: Check the Free Start and Current Limits

fileForensics launched on Product Hunt on November 4, 2026, with a free start, and like most tools in launch week, its exact plan limits can change quickly. Before you build a workflow around it — or recommend it to a client — check the current terms in the developer portal itself.

Here is a sensible way to do that check:

• Open the portal's pricing or plans page and note what the free tier includes: how many checks, which file types, and whether MCP and API access are included or paid.
• Test the free tier with real files for a week before deciding anything. A tool earns a place in your workflow by handling your actual files, not by its launch announcement.
• Do not pay for anything on day one. Learn the findings format, test the MCP connection, and confirm the check catches something real on a file you control. Pay only when the free tier is clearly too small for your volume.

Launch-week pricing is a moving target, so this guide deliberately does not quote numbers that may be outdated by the time you read it. The portal's own pages are the only reliable source for what is free today.

fileForensics AI vs Other Ways to Check a File

The table makes the trade-off clear. Extraction is fast but blind: it reads the words and trusts them. Manual checking is thorough but slow and needs an expert eye. fileForensics sits in the middle — fast like extraction, but it asks the trust question that extraction skips.

Frequently Asked Questions

Is fileForensics AI free?

fileForensics launched with a free start on Product Hunt on November 4, 2026. The free tier's exact limits — how many checks, which file types, and whether MCP and API access are included — can change during launch week, so check the developer portal's pricing page for the current terms before you rely on it. You can run your first checks without paying.

Does fileForensics give a trust score?

No, and that is deliberate. fileForensics does not give a black-box score. It returns findings with the evidence behind each one: what it found, where in the file it found it, and why it matters. You read the evidence and make the call, instead of trusting a number you cannot inspect.

Which AI tools does fileForensics work with?

It works through four routes: the developer portal in your browser, a hosted MCP server for Claude, Cursor, and Codex, a one-call API for apps and automations, and the fileAI platform itself. The portal is the best place to start; MCP is the best route when you want your AI agent to check files by itself.

Can fileForensics prove a file is fake?

No. It can show you evidence — signs of editing, AI-generated content, tampering, or internal contradictions — but evidence is not proof. A finding is a reason to investigate further: open the file, look at the spot the finding names, and verify with the source before you accuse anyone or reject a file outright.

Do I need to install anything to use fileForensics?

No. The developer portal runs in your browser with nothing to download. You only need extra setup if you want the MCP connection in Claude, Cursor, or Codex, or the API in your own app — and both of those start from the portal, following the steps in this guide.

What kinds of files can I check with fileForensics?

Documents, PDFs, images, and other common file types your AI works with. Upload the original file whenever possible, because screenshots, photos of screens, and re-saved copies lose the traces the check looks for. Scanned documents can be checked too, but expect more noise in the findings.

Final Thoughts

AI is only as trustworthy as the files you feed it. Until now, most of us handed documents to our AI assistants without a second thought — and the assistants, helpful as they are, never thought to ask whether the file itself was honest. fileForensics AI closes that gap with a simple idea: check the file first, then let the AI work.

Start in the developer portal with a file you trust, learn to read the findings, and then connect it to your agent with MCP so the check happens automatically. Keep the limits in mind — evidence, not proof — and pair the check with source verification whenever the stakes are high. That is the whole workflow, and it takes less time than fixing one bad decision made from a doctored file.

Explore More Free AI Tools

Want more free AI tools like fileForensics? Try our free tools at AIToolSteps — the AI Title Generator, Meta Description Generator, AI Rewriter, and Keyword Density Checker. They are free, beginner-friendly, and built to make your content work easier.