Skip to content
back
ART-02·CRITIQUE·28 December 2025·7 min

The nonsense of AI in marketing in 2026

By Camilo Rivera

Screenprint poster: a tall empty column drawn in a thin off-white line, with a small yellow disc resting at its foot
Generated poster · art direction by OSOM Labs

"AI will revolutionise your marketing." "Automate everything with AI." "10x your productivity." You read this everywhere. We use these tools every day. Here's what we actually observe, far from the promises.

0101

Promises vs reality

What they sell us:

  • "Generate 100 SEO articles in 1 click"
  • "AI analyses your data and finds insights"
  • "Create ad campaigns automatically"
  • "Replace your marketing team with AI"
  • What we observe:

  • Articles mass-produced to please Google are precisely what Google sets out to demote
  • AI doesn't find insights: it summarises what you give it
  • "Automatic" campaigns are generic and underperform
  • Replacing a team with a tool moves the work without removing it: someone still has to decide, check and correct
  • The problem: many AI marketing tools add an interface and a margin to a language model you can already use directly. Before paying for one more subscription, one question is enough: what does this tool do that the general-purpose assistant doesn't?

    0202

    The 3 categories of AI marketing tools

    After exploring dozens of tools, we see three categories. We describe them by use rather than by name: products change quickly, uses stay.

    1. The genuinely useful

  • General-purpose assistants (ChatGPT, Claude) for writing assistance, not replacement, and for brainstorming
  • Synthesis and organisation of information inside note-taking and documentation tools
  • Automation of simple repetitive tasks between the tools you already use
  • Assisted research that cites its sources, provided you open the sources
  • 2. The "it depends"

  • Image generation: useful for prototyping, risky for final output, because the rights remain unclear
  • Chatbots: well configured, useful; poorly configured, a disaster
  • Assisted SEO: suggestions, yes; automatic execution, no
  • 3. The nonsense

  • "Viral" post tools: posts "generated to go viral" are obvious from a mile away. It doesn't work.
  • "Predictive AI" you can't question: with what data? On what model? If the vendor can't answer, it's marketing on thin air.
  • "Replace your agency with AI": if that were true, agencies wouldn't exist any more. Still waiting.
  • 0303

    Why 100% AI content is a dead end

    Google said it as early as February 2023: appropriate use of AI is not against its guidelines, and its systems reward high-quality content however it is produced. In March 2024, it added a policy against content mass-produced to manipulate rankings, again no matter how it is created. So AI content isn't penalised per se. Mediocre content is, and a lot of AI content published as-is is mediocre.

    The problem: AI content published without human work is often:

  • Generic (the same structures, the same phrases like "let's dive in...")
  • Factually dubious (hallucinations published without checking)
  • Lacking real expertise (compilation of what already exists on the web)
  • Recognisable (yes, even by humans: the tics are easy to spot after a few articles)
  • What it means:

    Google doesn't punish AI, it punishes mediocrity at scale. Publishing hundreds of generated articles to fill a site is exactly the case its March 2024 policy targets.

    0404

    The real cost of AI (not the one they tell you)

    Visible costs:

  • A subscription to a general-purpose assistant
  • Specialised tools, often several
  • The API, if use becomes intensive
  • These are the only ones you're told about, and not the heaviest.

    Hidden costs:

  • Learning time: Learning to prompt effectively takes months
  • Verification time: Every AI output must be checked (otherwise: errors, hallucinations)
  • Correction time: Raw AI content is rarely usable as-is
  • Opportunity cost: Time spent tinkering with AI vs doing the work directly
  • Our observation: For simple, repetitive tasks, AI saves time. For complex tasks requiring expertise and nuance, it often wastes it.

    Learning time, on the other hand, can be shortened: that's what our Claude training is for.

    0505

    What actually works (our experience)

    In daily use, here's what we've found useful:

    It works:

  • Brainstorming and idea generation (starting point, never final result)
  • Reformulation and improvement of existing text: AI as a "second pair of eyes"
  • Research and information synthesis (with systematic verification)
  • Automation of repetitive technical tasks
  • Debugging and solving technical problems
  • It doesn't work:

  • Replacing industry expertise: an LLM doesn't know your market
  • Generating final content without human intervention: it always shows
  • "AI marketing strategy" without human strategy behind it: AI executes, it doesn't think
  • Business decisions based solely on AI: it's a tool, not an adviser
  • Our rule: If the AI output can be published as-is, it probably doesn't bring anything new.

    Sorting what can be delegated from what stays with you is also where our AI integration work starts.

    0606

    Our position

    We explore AI every day. We see its potential. We also see its limits.

    What we think:

    AI is an amplifier. It amplifies the skills of those who have them. It doesn't create skills in those who don't.

    An SME that uses AI without a clear strategy will produce more mediocre content, faster. An SME with real expertise can go further, more efficiently.

    The nonsense is promising that AI replaces thinking. It replaces nothing. It accelerates. And accelerating in the wrong direction is still going in the wrong direction.

    Key takeaways

    01

    Many AI marketing tools add an interface and a margin to a model you can already use.

    02

    Google doesn't penalise AI: it demotes mediocre content produced at scale, however it is written.

    03

    The real cost includes learning, verification and correction.

    04

    AI amplifies existing expertise; it doesn't replace it.

    05

    The nonsense: promising that AI does the work for you.

    Stack

    ChatGPTClaudeAI Marketing ToolsGoogle Search
    ai marketing · ai nonsense · ai tools · ai content · artificial intelligence limits · chatgpt marketing

    More articles in the journal

    We document everything we do.