11 Common Writing Mistakes (Plus AI Draft Red Flags to Fix)

Good writing usually does not fail because of one dramatic grammar error. More often, the problem is a collection of smaller issues: vague wording, unclear structure, repeated ideas, weak evidence, inconsistent tone, or sentences that make the reader work harder than necessary.

The fastest way to improve a draft is to identify what is making it harder to understand, then fix that problem directly. This guide covers 11 frequent writing problems with before-and-after examples, practical editing checks, and a three-pass revision method you can reuse on essays, reports, emails, blog posts, and professional documents.

This list has become more relevant, not less, now that most first drafts start with ChatGPT, Claude, or another AI tool. The same 11 problems still apply to a document whether a human typed every word or an AI generated the first pass. The difference is how each mistake shows up. A separate section further down maps these classic issues directly onto the tells that give away an unedited AI draft, so you can catch them faster.

Writing Mistakes at a Glance

Mistake 1: Writing Without a Clear Purpose

A sentence can be grammatically correct and still feel unnecessary. The problem is often that the writer has not decided what the document is supposed to accomplish.

Before writing, complete this sentence:

"After reading this, I want the reader to understand, decide, or do ______."

That answer becomes an editing filter. If a paragraph does not help the reader reach that goal, it may not belong in the document.

Example

Weak: "This report contains several important points about customer support that should be considered."

Clearer: "This report identifies three changes that could reduce customer-support response times."

The second version tells the reader what the document will actually do.

Mistake 2: Overusing Passive Voice

Passive voice is not automatically incorrect. The real question is whether it makes the sentence clearer or less clear.

Compare:

Passive: "The deadline was missed by the project team."

Active: "The project team missed the deadline."

The active version is shorter and makes responsibility immediately visible.

But passive voice can be useful when the actor is unknown, unimportant, or deliberately secondary.

For example:

"The samples were stored at 4°C before analysis."

Here, the procedure matters more than the identity of the person who stored the samples.

Purdue OWL similarly explains that passive constructions can be useful when the actor is unknown or when the writer wants to emphasize what receives the action. See Purdue's sentence-clarity guidance .

Do not chase an arbitrary percentage of passive sentences. Read each sentence and ask whether the actor needs to be clear.

Mistake 3: Run-On Sentences and Comma Splices

A run-on sentence occurs when independent clauses are joined without the punctuation or conjunction they need.

Problem: "The client approved the design, we started development the next day."

Both parts can stand alone as complete sentences. A comma alone does not correctly join them.

You could write:

• "The client approved the design. We started development the next day."
• "The client approved the design, so we started development the next day."
• "The client approved the design; we started development the next day."

Sentence length itself is not the problem. A long sentence can be clear, and a short sentence can be confusing. Focus on whether clauses are connected logically and punctuated correctly.

Mistake 4: Using Vague Words Instead of Specific Details

Words such as "good," "bad," "things," "very," and "a lot" are not forbidden. The problem starts when they replace information the reader actually needs.

Compare:

Vague: "The meeting went very well."

Specific: "The team approved the budget and assigned deadlines for all three project stages."

The second sentence lets the reader understand what "went well" means.

A practical specificity test

Circle words such as:

• good
• bad
• thing
• stuff
• many
• some
• very
• really

Do not delete them automatically. Ask whether a more precise detail would make the sentence more useful.

Mistake 5: Giving a Paragraph Too Many Jobs

A paragraph becomes difficult to follow when it begins with one point, moves to an unrelated example, introduces a second argument, and ends somewhere else entirely.

A useful paragraph usually has one main job.

Ask:

1. What is this paragraph trying to say?
2. What information supports that point?
3. Why does that information matter?
4. How does it connect to what comes next?

If you cannot summarize the paragraph's purpose in one short sentence, consider splitting it.

Example

Suppose a paragraph begins by explaining product pricing, then discusses customer complaints, then ends with employee training. Those may all be useful topics, but they probably need separate paragraphs.

Mistake 6: Repeating the Same Point

Repetition often enters a draft because the writer is thinking while writing. The same idea appears in the introduction, again in the middle, and then a third time with slightly different wording.

Consider:

"The new process saves time. It is faster than the old process. Employees can complete the task more quickly."

Those three sentences make almost the same point.

A tighter version might be:

"The new process reduced the number of steps employees need to complete the task."

When editing, search for paragraphs that answer the same question twice. Keep the version with the clearest explanation or strongest evidence.

If you need to verify the overall size of a draft after removing repetition, use the CountFlows Word Counter. Word count is a measurement, not a quality score.

Mistake 7: Letting the Tone Shift Without a Reason

Tone tells the reader what kind of relationship the document is trying to create.

A customer-support article may be conversational. A contract should be precise and formal. A university assignment follows the expectations of its discipline.

Problems appear when a document moves between styles without a reason.

For example:

"Our onboarding process is straightforward and takes about five minutes. Users shall thereafter ensure that all credentials remain confidential. Don't worry, though, it's super easy."

None of those tones is automatically wrong, but together they feel inconsistent.

Before editing, choose three words that describe the intended voice:

Examples: clear, professional, friendly

Then check whether each section still sounds like the same document.

Mistake 8: Making Claims Stronger Than the Evidence

Specific numbers can make writing more convincing, but only when they are real, relevant, and supported.

Avoid adding a percentage merely because it sounds authoritative.

Weak: "This method improves productivity by 40%."

If you do not have evidence for the figure, the number makes the sentence less trustworthy rather than more trustworthy.

Better options are:

• cite the actual source
• describe your own measured result accurately
• qualify the statement
• remove the claim if you cannot support it

Qualified: "The shorter workflow may reduce the time required for repetitive editing tasks because it removes two manual steps."

The same rule applies to words such as "always," "never," "everyone," "proves," and "guarantees." Strong claims need strong support.

Mistake 9: Writing Without Considering the Reader

The same information should not always be explained in the same way. A beginner may need definitions and examples. An experienced reader may want the answer immediately.

Before drafting, identify:

• who will read the document
• what they probably already know
• what they are trying to accomplish
• what question they are likely to ask next
• what information they do not need

This prevents two opposite problems: explaining obvious information for too long or assuming knowledge the reader does not have.

Example

A software developer reading API documentation may want parameters, examples, errors, and response formats. A first-time user of the same product may simply need to know which button to press.

Mistake 10: Adding Words That Do Not Add Information

Filler is not simply "long writing." A 3,000-word guide can be concise if every section answers a useful question. A 300-word paragraph can feel padded if it keeps repeating the same point.

Look for phrases such as:

• "It is important to note that..."
• "It should be mentioned that..."
• "In order to..."
• "Due to the fact that..."
• "At this point in time..."

They are not forbidden, but many sentences become cleaner without them.

Mistake 11: Trying to Edit Everything in One Pass

Proofreading for commas while also deciding whether an entire section belongs in the document is inefficient. Large structural decisions and small surface corrections require different kinds of attention.

This is why a second read often finds problems that were invisible during the first one.

There is no universal rule saying every document needs exactly two or three proofreading rounds. A short email and a research report do not need the same process.

A better approach is to give each editing pass a specific purpose.

AI Draft Red Flags: How These Mistakes Show Up in ChatGPT and Claude Text

"Writing" increasingly means editing a draft that started with ChatGPT, Claude, or another AI tool. The 11 mistakes above do not disappear when an AI writes the first version, they just wear a different disguise. Knowing the AI-specific version of each mistake makes it faster to spot during a sentence-level editing pass.

Why Pasted AI Text Looks "Off" in Google Docs or Word

Beyond tone and phrasing, ChatGPT and Claude output often carries formatting that was meant for a chat window, not a document. Markdown symbols such as ## for headings or ** for bold can paste in as literal characters instead of rendering, em dashes and curly quotes can arrive as the wrong character encoding, and invisible characters like non-breaking spaces can hide inside the text. All of this can make a document look unpolished even after the wording itself has been edited.

The practical fix is to clean the text before it reaches your document, not after you have already formatted around it. Running pasted AI content through the AI Text Cleaner strips out stray Markdown, smart punctuation, and hidden characters in one step. For the full breakdown of what causes this and how to fix it manually, see the guide on how to remove ChatGPT formatting from copied AI text.

Before-and-After Writing Examples

The CountFlows 3-Pass Editing Method

Instead of trying to fix an entire draft at once, review it in three layers.

Pass 1: Structure

Read only the title, headings, and first sentence of each paragraph. You should still be able to understand the path of the document.

Pass 2: Sentences

Read the body carefully. Check who is doing what, whether each sentence adds new information, and whether vague wording can become more specific.

The CountFlows Sentence Counter can help you measure how a draft is divided into sentences. Use the result as information, not as an automatic judgment of quality.

Pass 3: Surface

Check punctuation, spelling, capitalization, links, names, dates, numbers, headings, and formatting after the larger revisions are complete.

This order prevents you from spending time polishing a paragraph that you later decide to delete.

Should You Read Your Draft Aloud?

Reading aloud can be useful because it forces you to move through the text at a slower pace. Awkward rhythm, missing words, accidental repetition, and overly complicated sentences may become easier to notice.

It is not a replacement for structural editing or factual checking. Treat it as one additional review technique.

Final Writing Checklist

Before publishing or submitting a document, ask:

• Can I state the purpose of this document in one sentence?
• Does the opening reach the main point quickly?
• Does each paragraph have one clear job?
• Are sentences joined and punctuated correctly?
• Have vague words been replaced where specificity matters?
• Is passive voice being used intentionally?
• Have I removed repeated ideas?
• Does the tone remain appropriate for the reader?
• Can I support important factual claims?
• Have I qualified claims that are not universal?
• Have I removed filler without removing useful context?
• Have I checked spelling, punctuation, links, numbers, and names?
• If this started as an AI draft, have I checked for em dashes, stray Markdown, and stock openers?

If the document has a strict length requirement, use the Word Counter after the final edit to verify the total rather than estimating it from the number of pages.

FAQs

Why does ChatGPT text sound robotic?
Language models default to safe, predictable word choices and even sentence lengths. That combination reads as smooth but flat, which is why AI drafts often need a specificity and rhythm pass before publishing.

What are the most common signs of AI-generated text?
Frequent em dashes, "not only X, but also Y" phrasing, Markdown formatting like bold headers and bullet points in plain-text contexts, and a repeated rule-of-three pattern are among the most recognizable tells.

Is passive voice always a mistake, even in AI writing?
No. Passive voice is useful when the actor is unknown or unimportant. The problem is when it replaces a clear statement of who did what, which AI drafts do by default more often than human writers.

How do I remove ChatGPT formatting when I paste it into a document?
Run the text through the AI Text Cleaner before pasting, or clean it manually using the steps in how to remove ChatGPT formatting from copied AI text.

Do I still need to proofread a document after using an AI humanizer or cleaner tool?
Yes. Formatting and tone tools fix surface issues, but they do not verify facts, catch weak structure, or confirm the document matches its intended reader. A structure and sentence pass is still worth doing.

Bottom Line

Better writing does not come from banning passive voice, making every sentence short, or deleting every ordinary word. It comes from making deliberate choices.

Give the document a clear purpose. Make each paragraph do one job. Use specific language when details matter. Support claims that need evidence, and qualify statements when the evidence has limits.

Then edit in layers: structure first, sentences second, surface details last. That process catches more meaningful problems than simply running a spell checker and calling the draft finished. The same layered process works whether you wrote the first draft yourself or started from an AI-generated one, only the specific tells you are hunting for change.