AI website content vs manual content writing: What marketers need to know

For the past year, I’ve grappled a lot with how to weave generative AI into my workflow in a way that feels right to me. Here’s my personal golden rule for using AI in my writing process: If I wouldn’t ask a human to do it, I don't feel comfortable asking AI to do it. For instance, I wouldn’t ask my classmates in a writing course to write my essay for me. But I would ask them to workshop my draft with me and offer feedback, such as which parts they liked and which lines to cut. Because of that well-established human norm, the latter feels totally fair to ask an LLM, in my opinion.

Below is a content creation process I believe creates much better writing than AI-generated content (even AI-generated content that’s had human review).

Note: The creation process isn’t perfectly linear. I’ve outlined a step-by-step process for easy understanding below — but in reality, some steps happen concurrently. For example, I might conduct some research during the outlining stage, and I consider SEO/AEO optimization during the writing stage too.

Step 1 [Outline]: Human-First Foundation (Topic Selection, Content Brief, Outline)

While many recommend using AI for initial ideation and outlining, I disagree. I do not recommend getting AI involved with editorial decisions too early, if you can help it. Instead, I prefer to outline and draft first, then use AI as a research assistant or editor to help me polish the final draft. Why?

1. It preserves human editorial judgment. In a human-led, AI-assisted approach, you set the agenda by choosing which topics and keywords make it to the content calendar. And editorial decisions such as which products to review or which arguments to make in the article are reserved for human judgment — not AI.
2. It creates better outputs. I’ve found that starting with a human perspective creates stronger writing. Because LLMs are trained on the internet, their outputs are going to revert to the mean, meaning that if both you and your competitor prompt ChatGPT with, “Give me an outline for an article that targets the keyword ‘best mobile productivity apps,’” you’re both going to get very similar outputs, which is going to shape both of your articles into content that sounds rather generic.
   
   It’s better to come up with original insights and personal perspectives, then feed LLMs a prompt that is highly unique and detailed versus bland and vague. You’ll get much better outputs.

In step one of this hybrid human/AI writing process, start drafting the outline from your own brain. Bring in your lived experiences and perspective. We’re not looking for perfection or overt detail here; it can be very rough — just incomplete sentences and stream-of-consciousness ramblings. Later, you can polish and refine (using AI, if you want).

Step 2 [Research]: AI-Assisted Research and Organization

Next, start looking for stats, articles, and research you want to cite, as well as subject matter experts you’d like to interview to enrich your article with quotes. Line up your interviews and be sure to get permission to record and transcribe with AI.

How AI Can Assist with Research

Feed your human-led content brief, outline, research, and brand voice into your AI tools.

• Drop all of your links, PDFs, transcripts, and other research into NotebookLM to stay organized. This is better than having 30 tabs open or digging through your bookmarks to find that study you wanted to cite. Later, you’ll come back to NotebookLM for more insights and fact-checking.
• Need help finding research for your article? Consult ChatGPT, Claude, or your LLM of choice. ChatGPT’s deep research feature is especially useful here.
• It’s also helpful to chat back and forth with LLMs when trying to understand new or complex topics.

Step 3 [Write]: Human-Written First Draft

Now, for my favorite part: writing. Humans take the lead here, with AI as a thought partner and research/writing assistant.

This process is similar to that of Jana Rumberger, lead content strategist at HubSpot Media. “I do very little initial drafting with AI,” she tells me. “I do big, messy drafts. But I do a ton of perspective-based editing with [AI] engines that I've built.”

(We’ll revisit Rumberger’s perspective-based AI engines in step four, so keep reading for her fascinating take on how to use AI to break out of a writing rut.)

“I really get a lot from using AI in that way,” she continues, “and it's a lot faster for me than drafting something and then spending a lot of time editing the draft that AI has output. I feel like it allows you to maintain your authentic voice as a writer, while also making the process a lot faster.”

How to Use Generative AI in Content Marketing

Find out how to integrate AI tools into your content workflows.

• Understanding generative AI.
• Limitations of AI.
• Rolling out AI to your team.
• And more!

Download for Free Learn more Download for Free

Download Free

All fields are required.

You're all set!

Click this link to access this resource at any time.

Download Now

How AI Can Assist with the Writing Process

Quickly Finding and Incorporating Quotes, Research, and Citations

My favorite AI tool to assist with writing is NotebookLM. It’s a RAG-like tool, meaning it only pulls from the sources you add to it. This feels much more accurate and ethical to me, as I know that the outputs are from information I have vetted and trust, and I can easily verify and cite the information.

NotebookLM is a fantastic writing and research companion in that it always cites its sources  — you can click the citation to find the original source and ensure you credit it so you don’t accidentally plagiarize. Best tool ever.

Have a handful of interview transcripts from experts, but can’t remember who said what? Upload them to NotebookLM and ask something like, “Which expert mentioned their Instagram campaign that went viral?” NotebookLM will find the quote and link to it so you can locate it easily.

Pro tip: When using LLMs, prompt them to cite their sources. Then, fact-check those sources. If, for example, you’re doing research in ChatGPT and it mentions an idea, fact, or statistic, ask it to cite its sources by providing a link. Click through to that link and fact-check to ensure that ChatGPT is correct. Then, in your article, be sure to cite that original source.

Acting as a Thought Partner and Editor

I also loop in both Claude and ChatGPT as my thought partners and editors to provide feedback as I write, usually section by section. I bounce ideas back and forth between them, which strengthens my critical thinking and arguments.

Pro tip: LLMs are notorious for being “yes-men.” Because LLMs like ChatGPT and Claude are trained to fulfill the prompt you give them, that means they will try to give you what you seek — to the point of being people pleasers. This is great for stroking the ego, but terrible for getting feedback that improves your writing.

That’s why it’s crucial to prompt ChatGPT and Claude to think critically and push back on your writing.

Here’s a good example: As I was writing this section of the article, I realized I was rambling and my step-by-step process included too many steps and redundancies. So I asked Claude for feedback on tightening the language. I felt that “critique” was the wrong word to use, but Claude praised my word choice anyway:

Source

So, I specifically prompted Claude to “push back” on that feedback, and it thought harder and provided more critical feedback (note: “Extended thinking” was turned on, which forces Claude to think longer in the hopes of producing better answers. You can toggle this on for tougher questions and off for simpler ones.)

Source

Ultimately, Claude helped me land on a better one-word description of the iterative feedback process I use while writing with AI assistance: “revise.”

Source

Step 4 [Revise]: Soliciting AI Feedback

Alright, you’ve written your first draft. Now is when AI really shines — giving you critical feedback so you can revise your draft.

How AI Can Assist with the Revision Process

This is honestly where I’ve found AI to be the most useful. I ask ChatGPT or Claude to poke holes in my logic, find critical errors, and pick apart my weak arguments. I don’t always believe or use AI’s feedback, but it’s nice to be challenged by an LLM, as it often brings up points that I missed.

Friedman feels the same. “[AI] is really good at generating critiques that I can then scrutinize,” he says. “Because it is not directly working on the content, it is giving me things to think about that I can either reject or incorporate.

“And especially if it‘s trying to address a very particular persona, or reasoning through a complex topic that I’m not yet an expert on. It can help me fill in gaps, poke holes. Of course, if it‘s stuff that actually has to do with facts, then I’ll want to double-check that. I'm never going to take what it says for granted.”

Now, let’s revisit my colleague Jana Rumberger’s “perspective-driven AI engines.” How do they help strengthen her writing?

Here’s an example she gave me: She does a lot of reading and came across research on what makes something cool. Rumberger used that research to create a "cool engine.” This allows her to upload a draft to the AI, which will then analyze the draft and determine if it's cool or not. Fun, right?

Rumberger says that the mere act of getting a new perspective — even if it’s the wrong one — is useful.

“It gives me another framework to look at the writing through,” she explains, “and I think that that's one of the most challenging things about editing ... Oftentimes, getting that other perspective is what allows you to really make the turns that the article needs."

How to Use Generative AI in Content Marketing

Find out how to integrate AI tools into your content workflows.

• Understanding generative AI.
• Limitations of AI.
• Rolling out AI to your team.
• And more!

Download for Free Learn more Download for Free

Download Free

All fields are required.

You're all set!

Click this link to access this resource at any time.

Download Now

Step 5 [Optimize]: SEO/AEO

For much of my career, SEO was the star. Knowing it and implementing it well meant your content would shine. But now Answer Engine Optimization (AEO), also known as AI search, has eclipsed it (both are still important).

Once your writing is complete, you need to ensure optimization. This is something you should keep in mind during the writing, too, but doing a final check before publication is crucial.

SEO quick hits checklist:

• Identify your primary and secondary keywords. Sprinkle them naturally throughout your text.
• Include the primary keyword in H2s.
• Add alt text to all images that is descriptive, helpful, and, if possible, contains keywords.
• Link internally to relevant blog posts.

AEO quick hits checklist:

• Include semantic triples, which are subject + predicate + object statements. Example: “HubSpot helps marketers monitor campaign performance.”
• “Chunk” your content. “LLMs also index and retrieve content in ‘chunks,’” explains HubSpot's EN Blog Growth Manager Amanda Sellers in this blog post about AEO. “This means each paragraph or section in your piece of content should stand alone as a complete thought.”
• Saturate your chosen categories. Instead of narrowing it down to keywords, with AEO, you want to associate your brand with as many related topics and words as possible.
  
  Select topic clusters that you want to be known for. For example, a running brand might have the following clusters: “women's running shoes,” “men's running shoes,” “women's running apparel,” and "men's running apparel.”

Step 6 [Finalize]: Copyedits, Proofreading, Final Human Review

Your hybrid content is almost ready for publication. In this final step, you’ll use AI to help you with copyediting and proofreading, saving that final review for a set of human eyes.

How AI Can Assist

Grammar, Spelling, and Punctuation Checks

Grammarly excels at this, but I’ve found that ChatGPT and Claude sometimes catch mistakes that Grammarly missed.

Fact-Checking

Caveat: Be sure to do your own human-led fact-checking, and merely use AI as an extra layer (because humans can make mistakes, too!). Ask ChatGPT or Claude to check for factual errors. It’s useful to prompt the AI to take on the perspective of an expert. For example, if you’re writing about how to bake the perfect banana bread, ask it to think like an expert baker.

I also like to upload my article draft to the article’s specific notebook that I created in NotebookLM. Then I’ll give the AI the prompt of, “Review [article doc name]. Are all of the facts, info, and assertions supported by the sources?”

NotebookLM will then give you a line-by-line breakdown showing which sources support which parts of your article. It should also tell you if there’s an assertion in your article that is not supported by the sources. At that point, you can decide if you need to verify that fact again.

Brand Voice, Tone, and Style Consistency

You can also use AI tools to help ensure brand consistency. HubSpot’s Breeze AI Suite can learn and apply your unique brand voice to your selected content. Upload a writing sample so Breeze can analyze it.

Alternatively, you can create a custom GPT or Claude Project that preserves your brand voice, tone, and style guidelines. Upload the guidelines to the AI and have it check your article draft for violations.

If you’re using HubSpot’s Content Hub, the “finalize” stage is the perfect time to apply the brand voice feature to your writing.

Source

Next, be sure you read it over one last time before publication. All content, whether created with AI or not, benefits from human review. Read through your draft to make sure no errors slipped through.

Lastly, before hitting publish, be sure you’ve added a brief AI disclosure if you used AI to assist with your writing or if you’re publishing any AI-generated content.

How to Measure ROI for Hybrid Website Content

Measuring ROI for hybrid website content isn’t much different from how you were measuring ROI before — but there is an additional metric that’s all the rage now: brand visibility.

In AEO, brand visibility refers to how often your brand is mentioned and/or cited in LLMs (ChatGPT, Claude, Gemini, etc.). As Eric Carrell writes on LinkedIn, “A mention is your brand name appearing somewhere online. A citation is when AI quotes you as the source for its answer.”

Pro tip: Easily measure your brand AI visibility with HubSpot’s free AEO Grader. It will deliver a detailed report on your brand mentions, sentiment, and positioning against competitors in the AI search landscape.

Top metrics you should measure for your hybrid website content:

• Brand visibility
• Organic traffic
• Bounce rate
• Leads generated
• Conversion rate
• Revenue generated from website content

HubSpot Smart CRM makes it easy to measure ROI by connecting content to campaigns in Marketing Hub. Here’s how:

1. Log into your HubSpot account and go to Marketing > Campaigns.
2. Click the name of a campaign.
3. On the upper left, click Actions > Add assets.
4. In the left menu, click the asset type you want to add.
5. Check the box next to the asset(s) you want to add.
6. Click Save.

Source

Additionally, AI tools in Marketing Hub summarize and contextualize key content performance insights. You can even ask Breeze Assistant questions like, “Which blog post had the highest conversion rate?”

Should you disclose you’re using AI content, even if you’re editing it?

Yes, disclosing when you publish AI-generated content is the right thing to do. Plus, readers want AI disclosures. As a jumping-off point, here’s a useful template provided by Trusting News, based on its work and research:

“In this story, we used (AI/tool/description of tool) to help us (what AI/the tool did or helped you do). When using (AI/tool), we (fact-checked/made sure it met our ethical/accuracy standards) and (had a human check/review). Using this allowed us to (do more of x, go more in depth, provide content on more platforms, etc). Learn more about our approach to using AI (link to AI policy/AI ethics).”

The level and detail of disclosure often depend on your publication type and your audience’s expectations, but use the above as a starting point. Essentially, a good disclosure includes:

• What was used
• How it was used
• Why it was used

If you want to be even more concise, use a one-sentence disclosure and link to your full AI content process and policy page if readers want all the details.

Writing that does not have a byline, such as social media captions or landing page copy, has low to no expectations for disclosure. But bylined articles, such as those from news outlets or blogs, have high expectations of disclosure and transparency.

I personally don't expect disclosure when someone uses AI to assist with their writing (using it for research, grammar checks, feedback, etc.), but I do expect disclosure when they've used AI to generate the published words (the AI has the authorship in that case).

Note that I am specifically talking about disclosing the use of generative AI. In reality, publications have been using non-generative AI long before ChatGPT hit the scene, and there wasn’t an expectation to disclose.

For example, Grammarly, a popular grammar-checking tool, has used AI for its grammar and editing suggestions since it launched in 2009. And yet, that didn’t ruffle anyone’s feathers. But that’s because it was using different types of AI, and the type of AI that ChatGPT uses (generative AI) is meaningfully different. Generative AI is capable of generating entire written passages; that’s never been possible before and raises different ethical concerns.

Frequently Asked Questions About AI vs Manual Content

Will Google penalize AI‑assisted website content?

Google has stated that it does not penalize website content simply because it was created with AI; rather, Google penalizes any content (whether AI- or human-written) that is spammy or designed to game the system.

Here is the answer from Google Search Central:

“Appropriate use of AI or automation is not against our guidelines. This means that it is not used to generate content primarily to manipulate search rankings, which is against our spam policies.”

So, at least for Google, there doesn’t seem to be a difference in AI versus human content SEO value so long as it’s not spam. Whether you use AI or not, ensure your content is helpful and possesses E-E-A-T qualities.

My personal and professional advice on this, as someone who’s been writing for the web since 2009: Add concrete, detailed stories and examples from your personal experience. Show that you know what you’re talking about. Doing this alone will set you apart from 90% of the vague, unhelpful content that’s all over the internet.

Always have a human review and fact-check your content. Cite your sources. And always try to add that unique value.

When should you avoid AI for website content?

Pieces that need nuance, emotional depth, perspective, and a high level of factual accuracy and sensitivity should not be written by AI. You can, however, use AI to assist with research, feedback, and editing of the piece.

As always, make sure human judgment determines whether the AI outputs are used. For instance, you might ask Claude to help you find statistics relevant to an article you’re writing about heart disease. And Claude might give you some pretty compelling stats. But LLMs are known to hallucinate, so it’s up to you to find the source of those facts and ensure they’re legitimate and that the facts are actually, well, facts. (I like to ask ChatGPT and Claude to cite their sources so I can click those links and check. Both have commonly fabricated stats they provided me, and sometimes fabricated the links too.)

When I asked Rumberger when marketers should avoid using AI in their writing, she told me, "I‘m a big advocate for creative experimentation, so I wouldn’t say that there's anything that I would avoid.”

However, she added these cautionary words about using AI in the writing process: “Quality writing lies in the details, in being very thoughtful about how you execute on an idea and think about how each sentence expresses or leads up to an important idea. … If you outsource every step of that process to AI, it‘s going to be difficult to get an output that has any substance. And if you’re outputting something with no substance, then why should someone care about what you're sharing with them?”

How do I keep brand voice consistent when using AI?

To keep brand voice consistent when using AI, ensure you have centralized documentation that details your brand voice and tone and includes example copy. You can then share this documentation with your writing team and upload it to your AI tools.

For example, HubSpot's brand voice feature lets you use AI to tailor content to your brand's unique voice. Upload a writing sample or scan existing content to let the AI learn your brand voice. You can then apply it to social posts, blog posts, marketing emails, and more.

Additionally, Grammarly lets you create a style guide and set brand tones within the app; that way, it can offer customized suggestions when editing your content.

Lastly, always have humans review and refine AI-assisted copy to ensure it maintains your brand voice.

Should I disclose AI assistance on my site?

Yes. The public wants to know where it’s getting its information from. However, I understand it’s heavy-handed to disclose every little usage every single time.

For that reason, creating a blanket AI policy page with all the details is useful. Then, link to the AI policy in your footer. And for blog posts where you use AI, include a one- to two-sentence disclosure on your blog post and link to your full AI policy page; that way, it’s out there for the public to dive into at their convenience.

For example, we’ve published HubSpot Media’s core AI principles and how the HubSpot blog team uses AI. You'll also spot my AI disclosure at the bottom of this article.

AI Website Content vs Manual Content Writing: A Wild Time for Website Marketers

These days, I think, talk, and write a lot about AI. Considering the pros and cons of AI website content versus manual content is part of my job as a content marketer. I am constantly taking in new opinions and evolving my AI stance and approach. I might change my mind; this is just a hybrid process I’ve landed on for now.

Opinions on generative AI are on a spectrum. Either extreme (“don’t use generative AI ever!” or “use generative AI for everything!”) doesn’t seem reasonable to me.

Rather, I think it's more productive for content marketers who want to use AI to ask, “How can I use it to improve my writing, prove ROI to my organization, and still feel good about my work?”

This article was written by a human, but our team uses AI in our editorial process. Check out our full disclosure to learn more about how we use AI.

How to Use Generative AI in Content Marketing

Find out how to integrate AI tools into your content workflows.

• Understanding generative AI.
• Limitations of AI.
• Rolling out AI to your team.
• And more!

Download for Free Learn more Download for Free

Download Free

All fields are required.

You're all set!

Click this link to access this resource at any time.

Download Now