I use AI to generate content. I also know that AI-generated content often has patterns that Google rewards you for not having.

So I audit. Every piece of AI content goes through a five-step checklist before publishing. This is that checklist.

Step 1: Read it out loud (3 minutes)

Print the content or open it on your phone. Read it aloud. Not fast. Actually read it, like you're recording a YouTube video.

What you're listening for:

  • Sentences that sound unnatural. AI loves perfect cadence. Humans stumble.
  • Hedging language. "Arguably," "It could be said," "Perhaps" — these are AI fingerprints
  • Repetition of words or structures. AI does this more than humans.
  • Sections that feel abrupt or disconnected. Humans flow better.

    • If you hear problems, flag them. You'll fix them in Step 4.

    Real example of AI sentence I caught:

    "In today's fast-paced digital world, the importance of SEO cannot be overstated, as it enables businesses to gain visibility and drive meaningful engagement."

    Read that aloud. It's robotic. Humans don't write like that.

    Better version: "SEO works. It drives traffic. Without it, your competitors will beat you."

    Step 2: Skim for banned words (2 minutes)

    These are words that AI uses way more than humans:

  • delve, tapestry, landscape (metaphorically), robust (non-physical), comprehensive, nuanced, paradigm, synergy, holistic, streamline, cutting-edge, pivot, optimize (as a verb for non-technical things), reimagine, elevate (metaphorical), navigate (metaphorical), harness, foster, catalyst, leverage (verb), unpack, deep dive, at the end of the day, needless to say, it's important to note, moreover, furthermore

    • Find these words, delete them, replace with human alternatives.

    Real example from AI content I caught:

    "We need to leverage our digital landscape to catalyze meaningful engagement."

    Translation to human: "We need to use our website to reach people."

    Step 3: Check for evidence (5 minutes)

    AI generates plausible-sounding claims. Verify them.

    Real example from my content:

    AI wrote: "Studies show that 73% of users will abandon a website if it takes more than three seconds to load."

    I didn't recognize this stat. I searched for it. Couldn't find a credible source. Deleted the sentence.

    AI didn't make it up maliciously. It just generated something that sounded right based on its training data.

    Your checklist for Step 3:

  • Any statistic: search it (Google "73% page load" + your source)
  • Any quote: verify it's real
  • Any tool comparison: did you actually test the tools or did AI invent the comparison?
  • Any claim about what "most" people do: is there evidence?

    • Delete unverified claims. Add citations for important ones.

    Step 4: Add specificity (10 minutes)

    AI content is often correct but generic. Add specific details:

  • Real numbers from your experience
  • Specific tools you've tested
  • Specific examples of clients or situations
  • Specific mistakes you've made

    • Real example of AI content vs. edited content:

    AI version: "Keyword research tools are useful for identifying search opportunities."

    Edited version: "I use DataForSEO for keyword research. I spent $180/month on it last year. It's better than SEMrush for programmatic keyword analysis because the API is cleaner. I'd recommend it if you're doing API-based research. If you just need a dashboard, stick with SEMrush."

    The edited version is specific, opinionated, and useful. The AI version is ... fine.

    Step 5: Check the outline (5 minutes)

    AI often structures content in templates. Check if your actual outline makes sense.

    Common template AI uses: 1. Introduction with context 2. Three body sections 3. Conclusion that restates intro

    This works. It's boring, but it works.

    Better outline: 1. Start with a problem or insight (no setup) 2. Explain the context that matters 3. Show three specific things 4. End with a question or new thought (not a restatement)

    If your AI post follows the template too closely, restructure it.

    Real example of restructuring:

    AI version:

  • Intro: "Keyword research is important for SEO"
  • Body: Tools, methods, benefits
  • Conclusion: "Keyword research is important"

    • Restructured:

  • Intro: "I built an AI agent that does keyword research. It finds opportunities I'd miss manually."
  • Body: How it works, what it found, why it matters
  • Conclusion: "The real value isn't the speed. It's the consistency."

    • The speed round

    If I'm in a hurry (publishing 5 posts in a day), I do a faster version:

    1. Skim for banned words (1 min) 2. Check for one unverified claim (2 min) 3. Add two specific details (3 min) 4. Read the conclusion aloud (1 min)

    Total: 7 minutes. Catches 80% of the problems.

    What you're actually looking for

    The core audit question: Does this sound like it was written by someone who did the thing, or someone who knows about the thing?

    That distinction is everything for SEO.

    AI can write about keyword research. It can't write about your keyword research. The second is what ranks.

    So you're not auditing for "is this AI?" You're auditing for "does this sound like lived experience?"

    If it doesn't, you fix it in Step 4. You add the specificity that only someone who actually did the thing would know.

    That's the whole process.