Creating Content in a World of AI

6 min read

Two things are true at the same time right now, and a lot of businesses only seem to have absorbed one of them. First: Google has been explicit, including in its own published guidance, that it does not penalise content simply because AI was involved in producing it. Second: Google’s March 2026 core update specifically targeted synthetic, low-value “AI-slop” content, with a meaningful share of previously top-ranking pages dropping out for offering nothing beyond a summary of what already ranked. The tool isn’t the problem. What the tool was used to produce is the problem.

Google Doesn't Ban AI — It Bans Low Value

Google’s published guidance on generative AI content is fairly direct: the policy targets low-value, unoriginal content created at scale to manipulate rankings, regardless of how it was produced. AI-assisted content is explicitly permitted when it’s helpful, accurate, and genuinely created for people. What gets penalised is using AI to generate large volumes of pages that don’t add anything a reader couldn’t already find elsewhere — Google refers to this as scaled content abuse, and it applies exactly the same way whether the thin content was written by an underpaid freelancer or generated by a script.

The "Information Gain" Bar

The clearest signal from the March 2026 core update is the emphasis on what’s sometimes called information gain — whether a piece of content adds something genuinely new to what’s already available: original data, first-hand testing, a specific case example, an expert perspective not found elsewhere. Content that simply reorganises and rephrases the existing top ten results, however well written, increasingly fails to rank, because there’s nothing in it a search engine (or a person) couldn’t get from the sources it was based on.

What AI Overviews Have Done to Search Behaviour

AI Overviews — Google’s AI-generated summaries shown above traditional results — now appear on a substantial and growing share of search queries, and their presence measurably reduces click-through to the traditional organic results below them, particularly for top-ranking pages. That’s a real headwind for a content strategy built purely around ranking position. The flip side: content and brands that do get cited inside an AI Overview can see a meaningful lift in click-through from that citation, because it acts as a visible, credible entry point into your website rather than just one blue link among ten.

This changes the practical goal of content strategy somewhat — from purely “rank in position one” to “be the source an AI Overview (or an AI assistant like ChatGPT or Perplexity) actually chooses to cite,” which tends to reward the same things information gain rewards: clear, well-sourced, genuinely original content over generic summaries.

What 'Non-Commodity' Content Actually Looks Like

Original data, surveys or analysis that didn’t exist anywhere else before you published it
First-hand experience — screenshots, testing, results, a specific project or client example
A clearly identified, credible author with relevant expertise, rather than an anonymous byline
A genuinely distinct point of view, rather than a balanced restatement of what every competitor already says
Direct answers to specific, real questions your customers actually ask, not generic top-of-funnel topics

Trust, Written by a Real Person

Trustworthiness — the T in E-E-A-T — increasingly depends on whether a reader (and, by extension, Google’s quality systems) can identify who actually wrote a piece of content and why they’re qualified to. Content published under a real, named author with genuine, verifiable expertise in the subject tends to earn more trust than anonymous or generically-bylined content, regardless of how the first draft was produced. This is one of the more overlooked practical takeaways from the current environment: something as simple as a proper author bio, consistently used across a site, is a meaningful trust signal that costs almost nothing to implement.

Where AI Genuinely Helps in a Content Workflow

Used well, AI is a legitimate production accelerant: drafting structure, summarising research, generating variations to test, handling first-pass editing, or helping a subject-matter expert who isn’t naturally a writer get their knowledge onto the page faster. None of that is the problem Google’s updates are targeting. The problem is treating AI output as the finished product, rather than the first draft of one.

Where It Can't Replace a Human

AI can’t have actually used your product, sat in the client meeting, run the campaign, or made the judgment call your business is known for. It can’t originate a genuinely new opinion informed by real experience, and it can’t be held accountable, by name, for whether the advice in an article is actually sound — which is precisely the trust signal Google’s quality guidelines are built around. Every piece of AI-assisted content we produce goes through human review specifically to add the things AI structurally cannot: real experience, a verified fact-check, and editorial judgment about whether it’s actually worth publishing.

A Quick Example

Two businesses each use AI to help produce a service page. The first prompts a generic AI tool for “content about pest control services” and publishes the output with minimal changes — the result reads like a slightly reworded version of every other pest control page on the internet, because that’s structurally what a general-purpose AI model tends to produce when given a generic prompt. The second uses AI to draft an initial structure, then has an actual technician review and rewrite it with specific detail: the exact treatment process used, local pest issues specific to the region, and a couple of real case examples. The second page has information gain. The first one doesn’t, and is exactly the kind of content the March 2026 core update was built to filter out.

Common AI Content Mistakes to Avoid

Publishing AI output with no human fact-check, especially for anything involving specific claims, prices or regulations
Using AI to generate large volumes of near-identical pages targeting slightly different keyword variations
Leaving content unattributed to a real, credible author when expertise and trust are exactly what’s being evaluated
Treating AI-assisted content as a shortcut to skip research, rather than a tool to speed up production of well-researched content
Ignoring how the content would read to a genuinely knowledgeable expert in the topic — if it feels generic to them, it will read that way to Google’s quality systems too

Our Process for Using AI Responsibly

We use AI as a production tool within a human-led process: research and structure can be accelerated by AI, but the point of view, the examples, and the final review are done by people who actually understand the client’s business and industry. Nothing gets published without checking it clears the same bar we’d apply to any other content — does this say something genuinely useful that a reader (or an AI system evaluating what to cite) couldn’t already get from the top five existing results on the topic.

If your content strategy right now is producing a high volume of generic articles and seeing rankings stall or decline, the information gain bar is very likely part of why. Talk to Medium Marketing about a content strategy built around genuine expertise, not volume.

About

Ethan Caldwell

Reflective Blogger

Dinesh Arasaratnam

Fractional Leader

Dinesh Arasaratnam is the founder of Medium Marketing, a performance marketing agency helping mid-to-large businesses drive growth through strategy, systems, and fractional CMO support. He has advised clients across sectors like energy, eCommerce, SaaS, and education, with a focus on data-driven decision-making and marketing efficiency.

contact us

Lorem ipsum dolor sit amet. Ut laboriosam atque cum porro neque et distinctio explicabo.

Lorem ipsum dolor sit amet. Ut laboriosam atque cum porro neque et distinctio explicabo et voluptas natus non quis obcaecati. Id animi eveniet ut Quis iure rem nisi magni qui quidem facilis aut omnis voluptas.