How To Write Valuable AI Content That Builds Your Credibility (Humanizing Isn’t Enough)

TL;DR: How to Humanize AI Content Without Losing Your Credibility

The problem: Most experts now write with AI, and readers can spot unedited AI content fast. On LinkedIn, users can now flag a post as “Seems like AI slop.” Once a reader suspects a machine wrote your post, they start doubting you know the subject.

What Google says: Google doesn’t penalize AI use by itself. Its quality raters give the lowest rating to pages that are almost entirely AI generated with little effort, originality or added value.

What readers and search engines reward: Knowledge gain. That means experience, client stories, first-hand data, surveys, experiments, opinions and feelings that a reader can’t find anywhere else.

The fix: Start with your own brain dump and let AI structure and polish it. If you start from AI’s draft, you end up fine-tuning its material instead of yours.

Make it a system: Build AI skills with a banned-word list, platform rules drawn from your best posts and mandatory elements such as data, a position or a client story.

Tools that help: Claude or ChatGPT for structuring, Descript for turning voice notes into text and NotebookLM for drafting from your own transcripts.

In July 2026, LinkedIn put a new option in the menu on every post. Right next to “Not interested” and “Report post,” members can now tap “Seems like AI slop.” The same week, Originality.ai published a check of 5,000 long LinkedIn posts and classified 81.2% of them as likely AI, up from roughly half in late 2024.

I use AI every day in my own business, and most of our audience does too. A creator survey we cover in our elearning statistics found that 57.3% of creators use AI daily. The tools aren’t the problem. The problem starts when an expert publishes AI’s first draft under their own name.

Once a prospect suspects a machine wrote your post, they stop asking whether your advice is good and start asking whether you know any of this at all.

For a coach, consultant or course seller, that doubt costs more than a few likes.

This guide covers why AI-sounding content hurts experts in particular, what Google says about it and the workflow we use at Learning Revolution to make AI work on our own raw material.

You’ll also get a starter banned-word list, a way to build a LinkedIn writing skill from your own posts and a checklist to run before you publish.

Note: This page contains affiliate links. If you purchase through links on this page, we may earn a commission at no additional cost to you.

Why AI-Sounding Content Costs Experts More Than Anyone

An expert business sells judgment. People hire a coach, buy a course or book a consultant because they believe that person knows something they don’t. Generic content sends the opposite signal, because it reads like the average of everything already online.

Research backs this up. In a 2025 study of 261 participants, people rated writers lower on trustworthiness, competence, caring and likability as the share of disclosed AI writing went up.

They were also less willing to interact with that writer again. The drop was sharpest for interactive, relationship-building writing, which is exactly what a LinkedIn post or a newsletter is.

I’ve argued for a while that AI is moving the value in expert work from labor to leverage. Drafting, summarizing and formatting are getting cheap.

Judgment, synthesis and earned credibility are getting more valuable, which is why I keep coming back to authority assets instead of more content.

What readers notice	What it signals	What to do instead
Stock words such as “delve,” “landscape,” “unlock” and “game-changer”	Nobody chose these words. A model did	Use the plain word you would say out loud
“It’s not just X, it’s Y” and other contrast formulas	A template, applied to whatever topic came in	State your point directly, then back it up
Three-item lists and evenly sized paragraphs everywhere	Writing on autopilot	Let the content decide the length and the count
Advice with no names, numbers or places	The writer has no direct experience	Add one specific from your own work
A hook line, then one-sentence paragraphs and emoji bullets	A LinkedIn formula copied from thousands of posts	Write the way you talk to a client
Balanced “on one hand, on the other hand” with no conclusion	The writer is afraid to take a position	Say what you think and why
Confident claims that turn out to be wrong	Nobody checked the draft	Fact-check every number, name and date

The LinkedIn AI Content Problem | Where Your Credibility Gets Judged First

For most coaches, consultants and course creators, LinkedIn is where prospects first size up your expertise. It’s also where AI-written posts have become impossible to miss.

Our LinkedIn stats roundup shows why the platform matters so much for B2B experts, and the 81.2% figure shows how crowded it now is with machine-written posts.

LinkedIn has started pushing back. According to TechCrunch, the July 2026 update did four things.

• Added a “Seems like AI slop” option to the three-dot menu on posts.
• Introduced new classifiers to reduce AI slop in suggested content from people outside a member’s network.
• Started privately flagging posts in members’ dashboards when heavy AI use makes them seem inauthentic.
• Replaced the “enhance your post” AI writer with a proofreading tool built to keep the member’s own voice.

LinkedIn’s chief product officer, Hari Srinivasan, put it plainly. “People come to LinkedIn to connect with real people and share their real perspectives, ideas, and expertise.”

Think about who sees your posts through suggested content. It’s people outside your network, the ones who don’t know you yet and haven’t decided whether you’re worth following. Those are the prospects a coach or course creator most needs to reach, and they’re the audience LinkedIn now filters hardest.

The posts that work share one trait.

Each one is built around something only the writer has, such as a conversation with a client, a number from their own business or a mistake they made and what it taught them. We’ll build exactly that kind of post later in this guide.

What Google Actually Says About AI Content

A lot of SEO advice claims Google punishes AI content. The evidence doesn’t support that. Google penalizes low-effort content at scale, and unedited AI makes low-effort content very easy to produce.

Ahrefs analyzed 600,000 top-ranking pages and found that 86.5% contained some AI-generated content. The correlation between AI share and ranking position was 0.011, which is effectively zero. Purely AI pages made up only 4.6% of the results, though, and they rarely reached the top spot.

So the real line runs between content that adds something and content that doesn’t. Here’s where Google draws it in its own documents.

Source	What it says	What it means for you
Search Quality Rater Guidelines, section 4.6.6	Raters give the Lowest rating when “all or almost all” of the main content is AI generated with “little to no effort, little to no originality, and little to no added value”	Using AI is fine. Publishing it untouched is the risk
Rater Guidelines, section 4.6.5	Scaled content abuse means “creating an abundance of content with little effort or originality with no editing or manual curation”	Pumping out dozens of lightly edited AI posts can hurt the whole site
Guidance on generative AI content	Generative models “predict a likely sequence of words” and site owners should fact-check AI content before publishing	Every claim, stat and name needs a human check
Helpful content guidance	Asks whether content provides “original information, reporting, research, or analysis” and says attribution “doesn’t replace the need for original effort”	Citing others isn’t enough. Add something of your own

Google also holds a patent on scoring information gain, meaning how much new information a page adds beyond what a reader has already seen. A patent doesn’t prove Google uses it in ranking, and Google’s statement on the March 2026 core update named no targets. Still, the direction is consistent across everything Google publishes.

The same logic applies to AI Overviews and answer engines. They summarize what’s common and cite pages that add something specific. A page that repeats the consensus gives them nothing to cite you for.

Knowledge Gain | The Ingredients AI Can’t Make Up

Knowledge gain is what a reader learns from you that they couldn’t get from the top ten search results or a quick chat with an AI model. It works because both audiences reward it.

Readers remember the specific story, and search engines and answer engines have a reason to rank or cite the page that holds it.

A language model can only predict what’s typical. It can’t know what happened on your client call last Tuesday, what your last launch earned or which lesson in your course students keep getting stuck on.

Those are the ingredients to build around.

Ingredient	What it looks like in an expert business	Where to find it
Personal experience	What you tried, what broke and what you’d do differently	Your calendar, project notes, old emails
Client or student stories	A before and after from someone you helped, shared with permission	Coaching notes, testimonials, support tickets
First-hand data	Your own open rates, conversion rates, refund rates or enrollment numbers	Your email platform, course platform and analytics
Surveys	Questions you asked your own audience, in their own words	A short form sent to your list
Experiments	Something you tested and measured, such as two price points or two webinar titles	Your own tests
Expert interviews	Quotes and insights from people you spoke with directly	Podcast recordings, calls, event sessions
A clear position	What you think the common advice gets wrong, and why	Your gut reaction to what everyone else says
Feelings and reactions	What frustrated, surprised or worried you along the way	Your memory. AI has none of it
Audience questions	The questions people ask you again and again	Inbox replies, comments, sales calls

Here’s a real example from our own files.

When we surveyed Learning Revolution readers, most respondents had more than ten years of experience in their field. Asked about their biggest hurdle, one wrote, “Just getting what’s in my head into a digitally sellable product.” Another named “the time required in taking my knowledge of a subject and constructing it into a good quality course.”

No AI model could have produced those lines, and they tell us more about our audience than any keyword tool. That’s the kind of material to feed AI, and it’s the kind of material that makes audience research worth the effort.

How To Create Valuable Content With AI (Start With Your Brain Dump, Not AI’s Draft)

The single biggest decision in an AI writing workflow is who supplies the raw material. If you ask AI to write a post and then edit it, you’re polishing the model’s ideas, structure and examples. You can spend an hour on that draft and still publish something that started as the average of the internet.

Flip the order. Give AI your own unfiltered thinking first, then ask it to organize and tighten what you gave it. Now every idea, example and opinion in the final piece came from you, and AI did the work it’s good at.

	AI-first workflow	Human-first workflow
Starting point	A prompt such as “write a post about pricing a course”	Your notes, voice memo or transcript on pricing a course
Whose ideas	The model’s best guess at typical advice	Yours, including the odd and specific ones
Examples	Generic or invented	Real, from your work
Voice	The model’s default voice	Close to how you talk
Editing effort	Spent removing AI patterns and adding substance	Spent tightening and checking
Risk of errors	High, because the model fills gaps with plausible guesses	Lower, because you supplied the facts
Knowledge gain	Low	High

I do a version of this with my own podcast archive.

On Leading Learning Podcast episode 436, I described how we’ve had more than 500 conversations on the show. When I use AI on that material, I start by asking what the model already knows about a topic and how to make it understand the outcomes I want. Only then do I upload transcripts.

As I said on that episode, “If you’re thoughtful about how you’re engaging with the AI, with the data you have available, I’m amazed at what you can actually get out of it.” The data I have available is the point. The model works on 500 real conversations instead of making something up.

Here’s how to run a human-first workflow on any piece of content.

Step 1: Dump Everything You Know, Messy Is Fine

Talk or type for 10 to 20 minutes without editing. Cover what you believe, what you’ve seen, the stories that come to mind, the mistakes you made and the questions clients ask. Many experts find talking easier than writing, so record a voice memo or a call with a colleague and transcribe it with a tool such as Descript.

Step 2: Add Your Proof

Attach the evidence you already have. That might be a screenshot of your sales numbers, a survey result, a client email you have permission to quote, a podcast transcript or a link to a study you trust. Anything you add here is something AI won’t need to guess.

Step 3: Ask AI to Structure, Not to Add

Give the model your dump and proof, and set firm limits on what it may do. A prompt along these lines works well.

Example prompt
Here is my raw brain dump and supporting material on [topic] for [audience]. Organize it into an outline for a [LinkedIn post / article / email]. Use only the ideas, examples and numbers I gave you. Don’t add statistics, stories or claims of your own. Where the outline has a gap, write it as a question for me instead of filling it.

Step 4: Let AI Interview You

Ask the model to read your outline and list the questions a skeptical reader would ask. Answer them in your own words, by voice if that’s faster. This step pulls out the details you know so well that you forgot to mention them, and those details often become the best part of the piece.

Step 5: Draft From Your Material, Then Edit Line by Line

Now let AI draft from the outline and your answers. Read every sentence and ask whether you’d say it to a client across the table. Rewrite anything that sounds like a brochure, and cut anything that only repeats the point before it.

Step 6: Fact-Check and Run Your Filters

Check every number, name, date and quote against its source, as Google’s own guidance recommends. After that, run your banned-word list and platform rules. A reusable AI skill saves the most time at this stage, because it applies the same filters to every draft without you retyping them, and the skills section shows how to build one.

How We Use AI at Learning Revolution

We use AI across our content work, and we hold it to one rule. AI works on our material, not the other way around. This article is an example. It started as a long, messy brain dump from our content lead, covering the angle, the risks and the workflow, before AI touched it.

Here’s where our raw material comes from.

• Podcast transcripts: More than 500 Leading Learning conversations with learning leaders, which we can search and draft from.
• Expert interviews: Conversations we publish on the site, such as our interviews with Pathwright founder Paul Johnson, Jason Blumer on building a network for firm owners and Wayne Turmel on running great webinars.
• Our own research: Reader surveys for Learning Revolution and the annual Strategic Outlook report that Celisa Steele and I publish through Leading Learning.
• Reader replies: My newsletter regularly asks readers to reply and tell me how they’re using AI in their expertise-based business. Those replies become questions we answer on the site.
• Our own numbers and decisions: What has worked and failed in our business, from course launches to platform choices.

Our own survey data shows how common AI content work has become. In our 2026 Strategic Outlook, which surveyed 156 learning businesses, half of the AI-adopting respondents we asked already used AI for content creation or curation. That makes original material the thing that separates one learning business from the next.

AI also helps with research, as long as judgment stays with us. When I test a course idea with my 3M Model, I use Claude and Perplexity to pull Reddit and Quora discussions from the people I want to serve. AI finds the conversations, and I decide what they mean. Our course idea validator follows the same principle.

Build AI Skills That Enforce Your Standards

A skill is a saved set of instructions, examples and reference files that an AI tool applies every time you use it for a task. In Claude they’re called Skills and Projects. You’ll find custom GPTs and Projects in ChatGPT and Gems in Gemini. Building one turns your standards into something the AI follows by default instead of something you remember to type each time.

We run several, and each one does a different job.

Skill	What goes into it	What it prevents
Editing skill	A banned-word list, banned sentence patterns and rules on punctuation and paragraph length	Stock AI phrasing and formulas slipping into published work
Persona skill	The author’s real positions, frameworks, career history and past writing	Drafts that sound plausible but say things the author never would
Platform skills	10 to 20 of your best real posts for one channel, plus rules drawn from them	Every channel getting the same generic voice
Content-type skills	Your process for a review, a how-to guide or a sales email	Missing sections and thin research

Our persona skill for my own writing draws on my newsletter archive, my ebook Earned, Not Taught and the Strategic Outlook report. It holds one rule above the rest. If a draft needs a fact, quote or story that isn’t in the file, the AI flags the gap instead of inventing it. Invented experience is the fastest way to lose the credibility this whole effort is meant to protect.

Give It a Word List You Don’t Want

Every model has favorite words and sentence shapes. List the ones you never want to see, along with the plain word you’d use instead. Here’s a starter list drawn from the editing skill we use.

Avoid	Use instead
Leverage, harness, utilize	Use
Delve into, unpack, explore the nuances of	Look at, explain
Landscape, ecosystem, realm	Market, industry, field
Robust, seamless, scalable	Strong, smooth, grows with you, or a specific fact
Unlock, elevate, supercharge	Get, improve, or the actual result
Game-changer, transformative, pivotal	A number that shows the change
In today’s fast-paced world	Delete it
It’s not just X, it’s Y	State Y directly
Let’s dive in, here’s the thing	Start with the point
Moreover, furthermore, additionally	And, also, or nothing

Words are the easy part. Sentence patterns give AI away just as often, so list those too. Ours bans em dashes, semicolons, colons in the middle of a sentence, rhetorical questions answered straight away and runs of three short sentences in a row.

Models forget rules under pressure, especially in long pieces. We back our editing skill with a small checker script that scans every draft for those patterns before anything goes out. If you don’t want to code, ask your AI tool to check the finished draft against your list as a separate final step, in a fresh chat.

Add Mandatory Elements

A banned list removes the bad. Mandatory elements force in the good. Tell your skill that no piece is finished until it includes at least one item from each group that fits the format.

• Proof: A number from your business, a survey result or a cited study.
• Experience: Something you did, saw or got wrong.
• A person: A client story, a student question or an expert you spoke with.
• A position: What you think, stated plainly, including where you disagree with common advice.
• A next step: One clear thing the reader should do after reading.

If the AI can’t find one of these in your brain dump, it should ask you for it. That one rule does more to humanize AI content than any prompt trick, because it makes the gaps visible before you publish.

How to Build a LinkedIn Writing Skill From Your Own Posts

A platform skill teaches AI how you write on one channel. It works best when you build it from real samples, because your best posts already contain the rules. You only need to make them explicit. Here’s the process, using LinkedIn as the example.

• Collect 10 to 20 of your own posts. Pick ones that sound like you and got real conversation going, not only the ones that went viral. Leave out anything AI wrote.
• Ask AI to describe your patterns. Paste the posts into a fresh chat and ask it to describe how you open, how long your sentences and posts run, which words you repeat, how you use stories and how you end.
• Turn the description into rules. Edit the list until every rule matches how you want to write. Delete anything that describes a habit you’d like to drop.
• Add your banned list and mandatory elements. Every LinkedIn post should carry at least one real specific, such as a client moment, a number or a mistake.
• Save it as a skill, Project or custom GPT. Include the sample posts as reference files so the model can check its work against them.
• Test it on a brain dump. Give the skill a voice memo transcript and compare the draft with a post you wrote yourself. Adjust the rules where it drifts.

Use the same approach for your newsletter, sales emails and course descriptions. A newsletter skill would learn from your past issues. A sales skill would learn from the email sequences and course descriptions that converted.

The skill keeps your voice consistent, but it can’t decide what you stand for. That part comes from you, and our guides on thought leadership, personal branding and becoming known as an expert cover it in more depth.

Use Direct Response Copywriting Basics to Make It Land

Human-sounding content still needs to persuade. Direct response copywriting is the craft of writing that gets a specific reader to take a specific action, and its basics fix many of AI’s weakest habits. AI writes for everyone at once. Direct response writes for one person with one problem.

Celisa and I walked through one of the oldest frameworks on Leading Learning episode 144 on the AIDA formula. It still holds up, and so do these other principles.

Principle	What it means	The AI habit it fixes
One reader	Write to a single, specific person with a specific problem	Content aimed at “everyone” that lands with no one
PAS (problem, agitate, solve)	Name the problem, show what it costs, then offer the fix	Openings that warm up instead of starting
AIDA (attention, interest, desire, action)	Move the reader from noticing to acting	Posts that inform and then stop
Specificity	Use exact numbers, names and details	Vague claims such as “many experts” or “significant results”
Proof	Back every claim with evidence or a story	Confident statements with nothing behind them
One clear next step	Tell the reader exactly what to do	Endings that summarize instead of asking for action

Teach these to your AI skill as rules, then judge the draft yourself. A model can apply PAS. Only you know which problem keeps your clients up at night.

Before and After | What Humanizing Actually Looks Like

Here’s the difference in practice, using a LinkedIn post on the same lesson. The first version came from a one-line prompt. The second came from my own experience, which I’ve written about before in Earned, Not Taught.

AI-first version (from a prompt)
In today’s fast-paced digital landscape, building authority is crucial. It’s not just about what you know, it’s about how you show it. Here are 3 key tips:
1. Share valuable content
2. Be consistent
3. Engage with your audience
What’s your #1 tip for building authority? 👇

Human-first version (from a brain dump)
When Celisa and I started Tagoras, nobody had a reason to see us as LMS experts.
So we didn’t claim to be. We researched the LMS market and published what we found.
That research did what no bio line could. It gave people proof, and the authority followed.
What could you publish this quarter that proves what you know before you claim it?

The first version could come from anyone, and the list of tips could fit any topic. The second contains a real business, a real choice and a lesson the reader can test against their own situation. It also asks a question that’s hard to answer with a stock reply.

AI Writing Tools That Fit a Human-First Workflow (and What They Cost)

You don’t need a special humanizer app. The tools below support the workflow in this guide, and the right mix depends on whether you think better out loud, on paper or from source material. Prices update automatically from our pricing hub.

Tool	Best use in this workflow	Watch out for	Starts at
Claude	Structuring brain dumps, Skills and Projects that hold your rules and reference files	Still needs your facts and a final human read	Free plan, paid from $20/mo
ChatGPT	Custom GPTs, Projects and voice mode for talking through a brain dump	Strong default style that creeps back in long drafts	Free plan, paid from $8/mo
Gemini	Gems and drafting inside Google Docs	Same need for fact-checking	Free plan, paid from $4.99/mo
NotebookLM	Drafting and answering questions only from your own transcripts, notes and research	Limited to the sources you upload	Free plan, paid from $4.99/mo
Descript	Turning voice memos, calls and recordings into editable text	Transcripts need a quick check for names and terms	Free plan, paid from $24/mo
Grammarly	Final proofreading for spelling and grammar	Rewrite suggestions can flatten your voice	Free plan, paid from $30/mo
Jasper	Shared brand voice for teams with several writers	Priced for teams, more than most solo experts need	From $69/mo

For a solo coach or course creator, one general assistant plus a transcription tool covers almost everything here. Our full list of AI tools for course creators and our Descript review go deeper, and the Jasper review explains when a team platform is worth it. For more ideas on what to ask these tools, see our ChatGPT prompts guide.

Is “Undetectable” the Right Goal?

Much of the advice on this topic, including earlier versions of this article, focused on getting AI content past detectors. That’s the wrong target. Humanizer tools that swap words to fool a detector leave the substance untouched, so the content still says nothing a reader couldn’t get elsewhere.

Detectors themselves make mistakes in both directions, and a score tells you nothing about whether the content helps anyone. Use one, such as Originality.ai, as a rough smell test if you like. The real test is whether a client would learn something from the piece that they couldn’t get from a chatbot in ten seconds.

Should You Disclose AI Use?

Disclosure is a trade-off, and the research shows both sides. In Trusting News research, 94% of people wanted to know when AI was used, yet 42% said they were less likely to trust a story after seeing an AI disclosure. Another 91.5% said knowing whether a human reviewed the content was very important.

My take is to be honest about your process when AI played a real part, and to make clear that the ideas, experience and final review are yours. Google’s guidance points the same way for content where readers would reasonably expect to know how it was made. A human-first workflow makes that disclosure easy, because the substance really is yours.

A Pre-Publish Checklist for Human-First Content

Run through this list before anything goes out under your name. If you can’t tick most of the boxes, the piece isn’t ready.

• The core ideas came from my own brain dump, notes or transcripts.
• The piece contains at least one thing only I could have written, such as a story, a number or a position.
• Every statistic, name, date and quote has been checked against its source.
• No words or patterns from my banned list remain.
• I’d be comfortable saying every sentence out loud to a client.
• The opening gets to the point in the first two lines.
• The piece ends with one clear next step.
• Any client story is shared with permission.
• If AI played a real part, I’m comfortable explaining how I used it.

If you’re turning this kind of content into a course or a paid program, our guide to building an expertise-based business shows where it fits.

Frequently Asked Questions (FAQs)

Here are short answers to the questions we hear most about using AI for expert content.

What does it mean to humanize AI content?

Humanizing AI content means making sure the ideas, experience and judgment in a piece come from you, with AI used to organize and polish them. Swapping words or adjusting tone isn’t enough. A humanized piece contains specifics only you could supply, such as client stories, your own data and clear opinions.

Does Google penalize AI-generated content?

Google doesn’t penalize content for using AI. Its quality raters give the lowest rating to pages that are almost entirely AI generated with little effort, originality or added value, and it treats mass-produced, unedited content as spam. An Ahrefs study of 600,000 pages found no meaningful link between AI use and rankings.

Can people tell when a LinkedIn post was written by AI?

Often, yes. Readers spot stock phrases, formula hooks, one-line paragraphs and advice with no specifics. LinkedIn now lets members mark posts as “Seems like AI slop,” and Originality.ai classified 81.2% of long LinkedIn posts it checked in July 2026 as likely AI, so readers have plenty of practice.

What is a brain dump in AI writing?

A brain dump is your raw, unedited thinking on a topic, captured by typing or recording yourself talking. It includes beliefs, stories, mistakes, client questions and evidence. You give it to AI first and ask the model to structure it, so the final piece is built on your material instead of the model’s guesses.

How do I make AI write in my voice?

Build a skill, Project or custom GPT from 10 to 20 of your own best posts or emails. Ask the AI to describe your patterns, turn them into rules, add a banned-word list and require at least one real specific per piece. Then test it on a brain dump and adjust where it drifts.

Do AI humanizer and detector tools work?

Humanizer tools change wording to get past detectors, but they don’t add experience, data or opinions, so the content stays generic. Detectors make mistakes in both directions. Treat a detector score as a rough check at most, and judge content by whether it teaches readers something they can’t easily find elsewhere.

Should coaches and course creators disclose AI use?

Disclose when AI played a real part and readers would reasonably expect to know. Research from Trusting News found that 94% of people want AI use disclosed, though disclosure can lower trust. Making clear that the ideas, experience and final review are yours helps, and a human-first workflow makes that statement true.

Jeff Cobb, Founder of Learning Revolution

Jeff Cobb is an expert in online education and the business of adult lifelong learning. Over the past 20+ years he has built a thriving career based on that expertise – as an entrepreneur, a consultant, an author, and a speaker. Learning Revolution is a place where Jeff curates tips, insights, and resources to help you build a thriving expertise-based business. Learn more about Jeff Cobb here.