What Marketers Still Struggle With in AI and How To Get Better Results

AI has become a routine part of marketing work. Still, that doesn’t mean everyone’s getting the results they want.

In Brafton’s survey of marketing professionals, respondents reported persistent problems with generic content, inaccurate information, lengthy editing processes and a lack of genuine expertise in AI-generated output. At the same time, 61% of marketers using AI are learning how to use it without formal training.

We’re increasingly relying on AI for marketing work, but many of us are essentially still experimenting with how to use it effectively.

That presents some learning opportunities. So let’s dive into some of the leading concerns, because I have advice on how to address each one.

1. AI Content Is Too Generic

Of the 132 AI-using marketers surveyed, 87 identified thin or generic-sounding content as a concern. That makes it the most commonly reported issue by a considerable margin. If you’ve spent any amount of time using generic AI tools, you likely understand why.

AI-generated writing might be clear and technically correct, but it doesn’t necessarily offer anything readers couldn’t find in dozens of other articles. Oftentimes, the problem starts with the information we provide.

A prompt like „Write a blog about email marketing best practices“ gives AI very little direction about what the article should contribute, who it’s for or how it should differ from existing content.

And when most marketers are learning AI independently, figuring out how to provide better instructions often comes down to experimentation.

Try providing information and instructions that distinguish your content, including:

• Audience context: Who will read the content, what do they already know and what questions should it answer?
• Original information: What research, examples, customer insights or firsthand experience can support the article?
• Editorial direction: Which points deserve emphasis, and which commonly repeated explanations should be avoided?
• Brand voice: What writing examples and guidelines should influence the output?

Did you know: Both audience and brand briefs are set as standard reference materials for all contentmarketing.ai-generated content? That means you set them once, and all assets created under them follow their guidance in detail.

But if you’re working with another AI tool, here’s an example AI prompt to try:

Review this article for generic or low-value content. Identify statements that repeat common industry advice without adding useful context, evidence or examples. For each, explain what’s missing and recommend specific information that would make the section more valuable. Don’t invent supporting details.

This gives you a targeted list of opportunities to improve the article rather than simply asking AI to make it sound better.

2. AI Still Produces Inaccurate Information

Another 51 respondents expressed concerns about outdated or incorrect information. This is particularly important for marketers in regulated industries, or covering technical subjects, industry developments or anything involving statistics and research.

AI can provide a convincing explanation without having reliable evidence to support it. And when the output includes references or citations, it’s tempting to assume the information has already been verified. That’s not always the case.

Learning through trial and error has limitations here. Experimenting with prompts can help improve results, but it doesn’t necessarily teach someone how to evaluate sources or recognize inaccurate information.

Those are skills worth developing both in the context of AI for marketing and in general, whether through formal training or independent research.

I recognize that, for many, the whole point of AI was to streamline — and perhaps even combine — certain workflows. But there’s a case for keeping research and drafting separate.

For example, suppose you’re writing an article about recent changes to Google Search. Rather than asking AI to research the topic and produce a finished article in one step, you could:

1. Use a web-enabled AI tool to identify relevant Google announcements and documentation.
2. Review the sources and verify important claims.
3. Provide the approved findings to AI and ask it to develop an outline.
4. Generate content using those findings, with instructions not to introduce unsupported claims.

You can also ask AI to organize factual claims for verification.

Example AI prompt:

Review this draft and identify factual claims that require verification, including statistics, technical explanations and references to recent developments. Create a table showing each claim, its cited source and whether the source supports it. Flag anything that cannot be verified from the available material.

This doesn’t eliminate the need for human fact-checking. But it can make the process more systematic and help prioritize what needs attention.

3. Getting AI Content Ready To Publish Takes Too Long

Speed is one of the main reasons marketers use AI. But 46 survey respondents said getting AI output to the level of quality or authority their audiences expect takes too long. That essentially erases the initial benefit before you can even enjoy it.

Generating a first draft in two minutes is useful … until you have to spend another hour or two correcting problems.

Time savings depend on the entire production process, not just how quickly the initial draft is created.

Interestingly, Brafton’s research into how marketers are learning AI found that among respondents learning independently, the most frequently reported benefit of AI was being able to complete work faster. But there’s an opportunity to go further by improving the entire workflow, rather than individual steps.

One way to do that is to leverage AI throughout a structured process:

1. Research: Gather relevant information, sources and audience questions.
2. Outline: Organize the findings into sections with clear messaging points.
3. Draft: Generate content using the approved outline and supporting information.
4. Review: Identify repetition, unsupported claims, missing context and structural problems.
5. Optimize: Develop metadata, internal linking recommendations and promotional copy.

This gives you opportunities to correct problems before they manifest in a “finished” AI-generated article, and then trying to backtrack to where things went wrong.

Did you know: contentmarketing.ai deploys specialized agents that each tackle one of these tasks — rather than one agent that bounces between all of them at once. Plus, built-in checkpoints mean you can correct any inaccurate information, change direction or request new sources before it ever generates the article.

If you’re using another tool, it’s worth establishing quality expectations before drafting. For example, instruct AI to avoid repeating information between sections, support recommendations with examples and identify where expert input is required.

You can make those expectations reusable across assignments, too.

The broader point is to evaluate AI based on the quality and efficiency of the completed work rather than how quickly it can put together an initial draft.

4. AI Content Doesn’t Reflect Real Expertise

Another 43 respondents said AI-generated content doesn’t adequately reflect their expertise.

While related to the generic content struggle, it’s a bit different. You can improve the tone, structure and readability of an AI-generated article without necessarily introducing any actual new information.

For example, AI can explain common PPC optimization strategies. But it won’t automatically know what your paid media team learned from a recent campaign, which approaches failed or why a particular recommendation might not apply to your audience.

That information needs to come from the people doing the work. And for SEO and GEO, those original insights and firsthand experiences can be particularly valuable in helping content stand out.

The challenge is getting that expertise into your content without constantly pulling busy subject matter experts away from their work.

This is something contentmarketing.ai was built to address, and it’s reflected in several features:

• Knowledge Hub: Upload internal documents, technical resources, product information, interview transcripts and other proprietary materials. The AI will draw on that knowledge to inform content creation, rather than relying exclusively on generally available information.
• AI-generated SME quotes: Generate suggested expert commentary based on your brand guidelines, messaging and available knowledge. Instead of asking an expert to develop a quote from scratch, give them something to review, refine and approve.
• Asynchronous SME interviews: Use the Subject Matter Expert Interview workflow to generate questions, invite experts to respond on their own schedules and organize their insights for use in future content.

These applications help address a problem that better prompting alone won’t solve: AI doesn’t have access to your organization’s unique knowledge unless you provide it.

And importantly, they don’t remove experts from the process. Instead, these features make it easier to capture and apply their insights, while keeping people responsible for reviewing and approving what gets published.

5. Off-brand or Irrelevant Content

AI can generate technically accurate content that sounds perfectly acceptable to any average reader. But your audience isn’t the average reader — and your brand isn’t an average brand.

The problem often comes down to context. AI doesn’t automatically understand your company’s positioning, messaging priorities, target customers or what makes your products and services different. 

Give your AI tool context. Provide a comprehensive brand brief that outlines tone of voice, value propositions, audience personas and messaging guidelines. Include examples of content that represents your brand well, along with explanations of what makes those examples effective.

Here’s an example prompt that gets at context:

Review this draft against the attached brand guidelines and audience personas. Identify messaging that feels inconsistent with our positioning, language that doesn’t match our tone of voice and sections that aren’t relevant to our audience. Recommend specific revisions.

The more context AI has about your brand and audience, the better equipped it is to produce content that serves readers something true-to-brand and genuinely valuable.

Better AI Results Require Better Ways of Working

Generic content, inaccurate information, excessive editing and missing expertise are all legitimate concerns. But they’re also addressable via more deliberate prompting, better source material, structured workflows and critical review.

And while formal, structured training can help, marketers don’t need to wait for it to improve their AI skills. Most have already taken the matter into their own hands!

Note: This article was originally published on contentmarketing.ai.