What every tool does, the mistake people make with it, and what to do when the result comes back wrong. Knowing what a tool is for is rarely the problem — knowing why yours came back disappointing usually is.
Everything runs from the dashboard. Paste or type into the main box, pick a tool from the panel, press the button. That is the whole model — no projects to set up, no configuration before you can use anything.
The free word counter needs no account at all. Everything AI-assisted needs you signed in, because it costs money to run and the sign-in is how we know whose allowance it comes from.
What is included at each level is on the pricing page, which is kept current — figures quoted in documentation go stale and mislead people, so there are none here.
Paraphraser
Rewrites text you have already written, in a mode you choose. Seventeen of them, including five creative sub-genres and three built for professional documents.
What people get wrong: Pasting in a whole page. The paraphraser works on a passage — a paragraph, a clause, a sentence that came out badly. Give it two thousand words and it returns two thousand words of evenly adjusted prose, which is rarely what anybody wanted.
If the result disappoints: Shorten the selection and pick a more specific mode. Standard is the safest and the least interesting; almost any named mode gives you a stronger result because it has a direction to move in.
Summarizer
Reduces long text to its substance, with control over length and over whether you get prose or bullet points.
What people get wrong: Expecting it to know what matters to you. It summarises what the text emphasises, not what you were reading for.
If the result disappoints: If you need a particular angle, use the AI chat instead and say what you are looking for. "Summarise this with respect to the cost implications" is a question a summariser cannot be asked and an agent can.
Grammar check
Spelling, punctuation, agreement and clarity. Runs automatically a few seconds after you stop typing, or immediately when you press Analyze.
What people get wrong: Leaving it on while drafting. It interrupts the part of writing that most needs not to be interrupted.
If the result disappoints: Turn it off while the words are coming and on when you go back over them. One toggle, and it is the difference between a tool that helps and one you end up resenting.
Plagiarism scan
Checks your text against indexed web sources and reports matches with links.
What people get wrong: Running it the night before a deadline. A match found four hours out is a problem; the same match found two days out is a five-minute citation.
If the result disappoints: Run it on the draft, not the final. And read the matches — a quoted passage you cited correctly will still appear, and that is the system working rather than failing.
AI detection
Estimates how likely a passage is to have been machine-generated, as a percentage.
What people get wrong: Treating the number as a verdict. It measures statistical properties of the writing — predictability, sentence variation — not the origin of the text.
If the result disappoints: Use it as one signal among several. Plain, formal, carefully structured writing scores high whoever wrote it, which is why second-language writers get flagged disproportionately.
Humanizer
Rewrites text that reads as machine-generated so it sounds natural, with the meaning unchanged.
What people get wrong: Using it to get a lower detection score on something you did not write. It will often work and that is not what it is for.
If the result disappoints: Best on your own drafting that came out stiff — the paragraph where you were tired and everything came out in the same rhythm.
Translate
Translates between languages, keeping register rather than flattening everything to neutral.
What people get wrong: Assuming idiom survives. It handles meaning well and jokes badly, which is true of every translation system and most human translators.
If the result disappoints: For anything where tone carries weight, translate and then run the result through the paraphraser in the register you want.
AI chat
An open conversation rather than a tool with one job. Ask it anything about what you are writing.
What people get wrong: Asking it things a search engine answers. That is the most common first use and it is the least interesting one.
If the result disappoints: Give it a role and a task. "Read this as my department head and tell me what you would push back on" is the kind of question it is actually for — the AI agents page has eleven more.
Image generator
Generates an image from a description.
What people get wrong: Naming a subject and stopping. "Remote work" returns the same laptop-on-a-desk everybody else has, because a category has an average and the average is what comes back.
If the result disappoints: Subject, then action, then setting, then light, then style. And one detail that would only be there if the scene were real — a cold coffee, a chair pushed back.
Seventeen is more than anybody wants to read through, so the short version. Standard is the default and the least opinionated. Fluency is for text written quickly or in a second language. Formal and Casual move the register in either direction. Simple is for readers outside your field. Expand and Shorten do what they say.
Creative has five sub-genres — sci-fi, fantasy, horror, romance and literary — which are genuinely different instruments rather than five labels on one prompt.
Legal, Medical and Energy are built around what they must not change: carve-outs, doses, measurements, and the difference between what happened and why. They need ProMax, and the legal, medical and energy pages each show a worked example.
The browser extension works wherever you type on the web. Sign in once inside the extension — it does not inherit the session from the website, which is the single most common confusion about it.
The Windows desktop app reads selected text out of Word, PowerPoint, OneNote, Outlook and Scrivener, and most Windows applications that expose their text to the system. Highlight, pick a mode, and the rewrite happens in place.
It says sign up, and I am signed in
Almost always the extension or the desktop app rather than the website — each holds its own session. Sign in again inside the one that is complaining.
A mode I pay for is producing ordinary results
Legal, Medical and Energy need ProMax. If the account is Pro rather than ProMax they quietly fall back to Standard rather than refusing, so the output looks fine and is not the mode you chose.
The extension stopped responding after an update
Reload the page. An extension that updates while a tab is open leaves that tab talking to a version that no longer exists, and a reload is the whole fix.
The result is nothing like the examples
Usually the selection is too long. These work on a passage rather than a document — a paragraph gets a considered rewrite, a page gets an evenly adjusted one.
It cannot reach the service
If the website works and the tools do not, it is the connection to our servers rather than your internet. Some networks block them; trying on a phone hotspot confirms it in about ten seconds.
Something here that is wrong, or a question this does not answer? Tell us — this is a small project and it genuinely gets read.