The banned words list: a small lever for killing AI slop in your blog

If your AI-written posts all sound the same, a cleverer prompt probably won't fix it. A short list of words the writer is not allowed to use will do more.

That sounds too simple. It works because slop is mostly made of the same few dozen words and habits. Take those away and the writer has to say something real instead.

This post gives you a starter list and shows you how to add words from your own field. It also gives you a simple way to keep the list up to date as new offenders slip through.

Why a banned words list beats prompt tweaks

Most people fight slop by rewording the prompt. They add "sound natural" or "write like a human" or "be less formal". The next post comes back sounding much the same. So they tweak again.

A banned words list works better for four reasons.

It names a thing instead of a mood

"Sound natural" means something different to everyone, including the model. "Never write delve" means one thing. The writer can follow it, and you can see whether it did.

You can check it in seconds

A rule about tone needs judgement to review. A rule about a word is a yes or no. You can skim a post, or search it, and know straight away if a rule was broken. That makes review faster and less tiring.

It forces the sentence to change

When the easy word is gone, the lazy sentence around it usually goes too. Ban "unlock the power of" and the writer can't just swap in a synonym and carry on. It has to say what the thing actually does. That is where the quality comes from. The ban is small, but the rewrite it forces is not.

It adds up over time

Every rule you add is a small, lasting fix. Prompt tweaks drift. Someone rewrites the prompt, and last month's fix quietly disappears. A list only grows or gets pruned on purpose, so the gains stay.

What a list won't do

Be honest about its limits. A banned words list won't fix wrong facts, a weak outline or a post with no point. It is a small lever. It stops the most obvious tells, and that frees your review time for the harder problems.

A starter list of common AI tells

Here is a list to start from. You don't need all of it. Pick the ones that annoy you most.

Throat-clearing openers

These phrases fill the first line of a post without saying anything.

• In today's fast-paced world
• In the ever-evolving landscape of
• When it comes to
• It is important to note that
• It's worth mentioning that

Verbs that promise more than they deliver

These make a plain action sound grand.

• delve into
• unlock the power of
• unleash, harness, leverage
• elevate, empower, supercharge
• revolutionise, transform
• embark on a journey
• navigate the complexities of

Nouns and adjectives doing no work

These sound weighty and add nothing. Try deleting them and see if the sentence loses anything.

• landscape, realm, tapestry
• testament to, game-changer
• seamless, robust, cutting-edge
• pivotal, crucial, vibrant

Closers

These announce that the post is ending instead of just ending it.

• In conclusion
• In summary
• At the end of the day
• Ultimately, at the start of the last paragraph

Patterns, not just words

Some tells are shapes rather than words. You can ban these too.

• The em dash. Many writing rules ban it outright and ask for a comma, a colon or a second sentence instead.
• "It's not just X, it's Y." This setup appears again and again in AI posts.
• Lists of three adjectives in a row, like "fast, simple and powerful".
• Rhetorical questions as section openers, like "But what does this mean for you?"

Picking the offenders for your own field

The starter list catches the general tells. Your field has its own. These are often worse, because your readers see them everywhere and have learned to skip past them.

Read your own recent posts

Open your last few posts side by side. Highlight any phrase that shows up in more than one of them. Repeats are the clearest sign of a writer running on autopilot.

Look for your industry's buzzwords

Every field has words that were fresh once and now mean nothing. A few examples:

• Software: streamline your workflow, all-in-one solution, take it to the next level.
• Travel and hospitality: nestled, hidden gem, a feast for the senses.
• Property: boasts, stunning, must-see.
• Finance: financial freedom, peace of mind, take control of your future.

If a phrase could appear on any competitor's site without anyone noticing, it is a candidate.

Check the words your customers actually use

Read your support emails, sales notes or reviews. Your customers describe your product in their own words. Those are the words your posts should use. Anything your writer says that your customers never do is worth a second look. If customers say "cheap" and your posts say "cost-effective", you have found a rule.

Ban claims you can't back up

Words like best, leading, guaranteed and trusted by thousands make promises. If you can't prove them, ban them. This protects you as well as the prose.

Keep it short

A long list is hard to follow and hard to review. It can also push the writer into stiff, odd phrasing as it dodges dozens of words at once. Start with the handful that bother you most. Leave the rest until you catch them in a real post, and note those later catches in your website profile, as shown below.

Writing the rules so they stick

How you phrase a rule matters as much as which words it covers.

• Say what to do instead. "Don't use utilise" is fine. "Say use, never utilise" is better, because the writer knows where to go.
• Group near-synonyms. One rule like "Avoid leverage, harness and utilise. Say use" stops the writer swapping one for another.
• One idea per rule. A rule that covers five unrelated things is easy to half-follow.
• Ban the pattern when the words keep changing. If the writer keeps finding new ways to open with a question, ban opening with a question rather than chasing each wording.

A workflow for updating the list

Your list should grow from what you actually catch, not from what you guess. In RankMill it lives in two places. Your starting list becomes the project's writing rules when you create the project, and those can't be changed afterwards. Everything you catch later goes into the website profile, which you can edit at any time and which every agent reads. Here is a simple loop that works with that split.

1. Pick your starting list before you create the project. That is when it becomes the project's writing rules, so choose the words that bother you most and phrase each rule with an alternative, as above.
2. Review every post before it goes out. As you edit, note each change that is about a word or a phrase rather than a fact.
3. Sort the misses. Some are one-offs. Others are habits. If a phrase shows up in two posts, treat it as a habit.
4. Add each habit to the website profile. Keep a short section such as "Words we avoid" and list the habit there, with the word to use instead.
5. Fix the post in front of you. The profile change helps the next post. This one still needs the edit.
6. Watch for substitutes. Ban delve into and you may start seeing dive into. When a near-synonym appears, add it next to the original entry in the profile.
7. Tidy the profile section now and then. Read it every so often. Merge duplicates. Drop entries that make the writing awkward. An entry that causes more harm than the word it bans should go.

The loop is slow on purpose. Each entry earns its place by showing up in a real post.

Doing this in RankMill

In RankMill, your starting list lives in the project's writing rules. These are a plain list of sentences that every agent working on that project has to follow. You can pick from one-click suggestions for the common rules, or write your own. They are set when you create the project and can't be changed afterwards, so have your list ready before you start.

A few things make this handy:

• Rules belong to the project. If you look after several websites, each project gets its own list. A hotel blog and a software blog can ban completely different words.
• Later catches go in the website profile. Every agent reads the website profile, and you can edit it at any time. Add a short "Words we avoid" section and list each habit there. Edit it by hand in the Profile tab, or ask for an AI edit and accept or turn down the version it proposes.
• You can add rules for one post. When you write a post from an idea, you can add rules for that post on top of the project's own.
• Fixing a post you already have is quick. Open the post and tell the editor agent what to change, like "remove every use of seamless". Each change is made to the post straight away and shown as a card you can undo. For a one-word fix, just edit the post yourself in the block editor.

So the workflow above maps straight on. Review the post, fix it in the editor, and add the habit to your website profile so the next post doesn't repeat it.

Start with five

Don't build the perfect list today. Pick five items from the starter list that bother you most. Add them as writing rules when you create the project. Then review your next post and see what slips through. That goes in the "Words we avoid" section of your website profile.

A short list you keep up to date will do more for your posts than a long one you wrote once and forgot.