Strategies from Someone Who Lives It

It gives me a “second brain” and superhuman speed, and takes care of all the tedious tasks I dislike about my job.

Sure, maybe one day AI will replace my job. But right now? The real risk is not using it and falling behind everyone who is.

This guide shares everything I’ve learned from about two years of using AI in content marketing, often before my managers believed in it. I’ll show you when to lean into your human strengths and when to let AI take over.

AI is cheaper. About 4.7x cheaper than human-made content.

If you look at the distribution of content cost, you’ll see more AI users towards the left with the lower cost. Marketers who don’t use AI always pay more.

And AI makes everything cheaper, from blog posts to videos and whitepapers.

But is it actually any good, or are people just trying to cut down costs?

I’m going to give you first-hand examples of where AI is absolutely phenomenal, almost necessary. No theory, just tried and tested stuff I can’t imagine working without.

Come up with content ideas faster

For example, in my guides, I always suggest starting keyword research by letting AI come up with the first ideas (aka seed keywords). Don’t worry if you’re not sure what those are—AI knows, and it’s surprisingly good at finding them. You can ask any large language model (LLM) to help, or just use the AI inside Ahrefs’ Keywords Explorer to instantly generate thousands of keyword ideas from those starting points.

Pressure-test your ideas before you build them

You are my brutally honest content-strategy partner.
Your mission: tear apart my content idea so I can rebuild it stronger.
How to respond—use this exact structure:

Savage Snapshot (≤ 50 words): Your blunt first impression. No sugar-coating; if it sounds boring, derivative, or unclear, say so.

What Works (3 bullets max): The strongest elements worth keeping.

Critical Weak Spots:

Point-by-point flaws (logic gaps, weak angles, lack of differentiation, poor audience-fit, SEO/issues, data gaps, etc.).

For each flaw, suggest one specific fix or stronger alternative.

Impact Scorecard (1–10 each, with a one-line justification):

Originality
Audience Resonance
Search/Distribution Potential
Revenue/Conversion Potential

Kill-or-Keep Verdict: Choose KILL, REWORK, or GREENLIGHT and explain why.

One-Sentence Rewrite: If you marked REWORK or GREENLIGHT, rewrite the core idea/title in a punchier way.

Rules:
• Be ruthless but never vague—every critique needs a concrete fix.
• Assume my audience is {describe your audience briefly}.
• If you quote clichés (“ultimate guide,” etc.), propose fresher phrasing.
• Prefer brevity over fluff; no paragraphs longer than 4 lines.
• Do not recap my idea—just critique it.

Cut research time way down

I regularly crowdsource input for my articles. Sometimes, I use Help a B2B Reporter Out, and other times, I just post a request on LinkedIn. Within a few days, I usually get dozens of quotes from real people. And when it’s too much to sort through on my own… yep, I turn to AI. But not just any LLM. Most have context windows that are too small for this kind of task.

1. I use Apple Mail to save all HARO email notifications as PDFs.
2. Then I upload those PDFs to Google’s NotebookLM and give it a minute to process everything.
3. Once that’s done, I can ask ANY question about the content or filter for specific types of responses. For example, I might ask, “Show me the most unconventional tips from business owners, and note which file they came from.”
4. Once I’ve got the quotes and their sources, I switch to a different AI tool to help me incorporate the material into the article; NotebookLM is great for organizing and searching content, but it’s not the best for writing.

And I couldn’t recommend deep research features enough. You can find them in ChatGPT, Gemini, Claude, Perplexity, and probably a few others.

Deep research in LLMs means improving AI tools’ ability to find accurate, trustworthy, and detailed information. Instead of guessing based on what it learned during training, an LLM can now look up real-time facts from articles, websites, or documents. It can break down complex questions, search in steps, and use tools to double-check its answers.

For example, I used deep research on the data study that examined the correlation between the organic traffic of publicly traded companies on Nasdaq and their stock value.

I needed to check whether someone else had done similar research before me and what they had found. Gemini did that for me while I was having lunch.

And by the way, pretty much all of the number crunching in that article and a bunch of other data studies I did was all AI. I mean:

• Writing the code to get data from Ahrefs and Polygon via API.
• Merging two completely different databases (two files merged by a key that was in the third file).
• Calculating Spearman and Pearson correlations.
• Visualizing data.
• Creating a bot with all of my research data to make a database I could ask anything, anytime.

That said, the data study part of the article became much deeper and more interesting when I invited an actual data scientist to do their magic. AI doesn’t win every race yet, but more on that in a bit.

Get (optimized) outlines and first drafts out of the way

If you write with SEO in mind, like I do, using AI to outline or even draft your articles just makes sense. Google looks for consensus, and LLMs mirror it. Since they’re trained on the collective content of the web, they “naturally” pick up on the patterns and structures that tend to rank well.

Let’s take search intent for example. I remember when I had to find it manually to make sure my articles would match what the searchers expect. I’d open each of the top-ranking pages to see what it’s about and what subtopics I’d need to include in my article. Now Ahrefs’ AI Content Helper does that for me. And if the intent is mixed, it lets me choose what I want to tackle.

Then, the tool maps out exactly which topics and subtopics I need to cover in any new article—or shows what’s missing in an existing article—by analyzing what’s actually ranking in the top 10.

Automate parts of technical SEO

If you’re on a small content team, my guess is someone on that team handles technical SEO. And my other guess is that they don’t like it. Who would enjoy fixing dozens of meta descriptions and title issues, or asking their busy dev team to fix these?

SEO tools were quick to adopt AI to solve problems like this. So now, when Ahrefs flags a bunch of my articles for having too-long titles, I can fix them in no time. I just use AI to rewrite them all in one go and push the updates straight to the site, right from within Side Audit.

Make your content easier to read and brand guide compliant

AI tools can flag things like clunky sentences, broken grammar, awkward phrasing, and off-brand tone. It’s a useful way to clean things up and help your content perform better without rewriting everything.

For example, I made a bot for myself that’s trained on my past content to make everything sound like the publish-ready me.

I have another bot that generates WordPress shortcodes and meta descriptions so I don’t have to. All I need to do is upload the article.

What’s cool about it is that AI creates a nice, SEO-friendly URL slug for each subsection of the article.

Creating content reports

For example, you can use a bot like the one I built for the marketing analytics guide to run the numbers on your content marketing data. It’ll give you a full analysis and even some solid recommendations.

Another example. My editor, Ryan Law, used AI to write code from a description in natural language (aka vibe coding) for a Google Search Console reporting tool. He just describes what he wants, and the AI codes it.

And here’s the tool running, ready to create the report he needs:

Content repurposing; turning one piece of content into many

AI makes it easier to repurpose a blog post into an X thread, LinkedIn post, newsletter blurb, or video script. It can also help with scheduling and finding the best times to publish across platforms.

Notice I said, “It makes it easier,” but not that it does everything for you. I’m bullish on this tech, but I still wouldn’t trust it to post directly to my social channels on its own. That said, as a social media sidekick that never runs out of ideas or energy, it’s incredibly helpful.

Here’s an example of this: a content repurposing workflow designed in Zapier by Ryan. It waits for a new article on our blog, creates three types of LinkedIn posts and uploads them to Buffer for review.

And here’s a sample outcome.

Content marketing automation

Even before AI became a buzzword, content automation was already helping us escape the dull, repetitive parts of content management. But now, with AI agents, we’re moving far beyond just automating simple tasks. These tools are capable of handling even complex, creative work.

Imagine this: an AI agent notices when a new blog post goes live in your CMS. It then takes care of everything that follows: compressing images, generating metadata, scheduling social media teasers, updating your content calendar, and even logging each action for transparency. All of this happens automatically.

I was recently trying AirOps with some surprisingly good outcomes. The secret is that under the hood, it handles everything from topic research and search intent analysis to content creation, originality checks, and more through an AI automation workflow. Every part of it is customizable.

Now, AI agents are getting even smarter thanks to something called MCPs, short for Model Context Protocols. Think of MCPs as bridges that let AI talk to other apps more easily, without the need to write custom API code every time. They even allow one AI to communicate with another.

Here’s a quick example: I recently used Claude together with our new Ahrefs MCP. With just a prompt, it helped me discover fresh keyword opportunities for our blog and even brainstormed a few title ideas for each one.

It’s a fairly simple MCP AI agent, but I bet listing what it did there already makes you want to automate that stuff. It had to:

• See what the blog is about.
• Find the keywords that blog ranks for.
• Find new keywords excluding the ones it already ranks for.
• Brainstorm 5 titles for each keyword.
• Combine all findings and prepare a report.

Help with localizing your content

Our Head of International Marketing, Erik Sarissky, says the tech is not perfect yet, but it gets you at least half of the way in no time and makes human proofreading so much easier.

Now that ChatGPT can remember good responses, it can reuse accurate translations the next time it handles a similar localization task. Erik’s team uses that to build their own custom dictionary as they go.

Finish your thoughts

This is one of my favorite use cases and an example of how AI enhances your intelligence. OR, more accurately, brings you to results you would probably get anyway, but you’re there way quicker.

For example, a few moments ago, I got stuck trying to make my point sound good.

See what I mean?

You could write an entire opinion piece this way. You give AI the arc of the story or just a braindump of what you know or would like to say about the topic(the fun part), and it makes up the words in between (the boring part).