7 AI Mistakes That Wreck Your Traffic

And How Intelligent Content Marketers Can Get Better Results with AI-Generated Content

If you’ve only dabbled in ChatGPT, or you’re getting most of your information from Twitter or clickbait YouTube videos, you might be thinking you’re about to release a torrent of new blogs that will drive up your search ranking in record time.

Big mistake.

Yes, ChatGPT can help your SEO content production, but it still requires human judgment for key goals like:

  1. Achieving the content quality and expertise required for high search ranking
  2. Properly leveraging blogs as part of a well-designed content marketing strategy
  3. Researching keywords and using them in the right context

However, as “AI-powered” quickly becomes the next painfully overused tech buzzword after “data-driven”, it’s spawning all kinds of misguided advice to marketers. 

Only 10% of marketers use AI tools like ChatGPT for SEO content generation, but 58% plan to start using it in 2023. So to keep you from doing this the wrong way, let’s go over some common misconceptions, as well as how to get started using ChatGPT in your marketing. 

ChatGPT is a tool, not an employee

As the web floods with AI-generated text, more is required for your content to stand out.

Like any new hype cycle, it’s tempting to overstate the powers of new technology in its early stages. ChatGPT hasn’t replaced the need for results-oriented content writers, though it’s already transforming what the role will soon look like.

As one such results-oriented, tech-loving writer, I’ve tested ChatGPT and prior GPT models extensively since their respective releases. But while these tools have been exciting and helpful, they have big limitations.

ChatGPT isn’t a replacement for human content production, but a supplement. A bicycle will get you from A to B faster, but the bike’s not going to get there on its own. It needs you to pedal — and more importantly, to steer.

Yes, content marketers will be spending less time in “workhorse” writing mode when we have ChatGPT’s help, but that’s not an excuse to outsource your brain function to the robots just yet. The bar has been raised (or maybe lowered, if we’re playing limbo).

Less time writing means more time revising and improving content quality, and going the extra mile to make your content more useful, remarkable, or delightful to read. It gives you more bandwidth to build trust with your readers and establish your authority on a topic. 

Business leaders who fail to realize this will be left in the dust by those who do.

So let’s go over what ChatGPT can and can’t do for SEO, how using it wrong can hurt your search traffic, and how you can use it for a better marketing strategy.

ChatGPT-supported content still needs to meet SEO standards

When it comes to making content that ranks, Google cares about things like “E-E-A-T” (or “Experience, Expertise, Authority, and Trustworthiness”), along with metrics like Relevance (to a search) and On-Page Experience (like easy readability and fast load times). ChatGPT won’t do this by itself, so you need to be deliberate about how you use it.

According to Google Search Central’s recommendations on AI content:

However content is produced, those seeking success in Google Search should be looking to produce original, high-quality, people-first content demonstrating qualities E-E-A-T.

In other words, great content should:

  • Help users — that means high-quality, substantial content: no generic babble or fluff. The user should come away feeling satisfied with what they read; that they’ve learned enough to reach their goal.
  • Demonstrate expertise  — it’s informative, with original reporting, research, or analysis.
  • Be posted on an authoritative site — particularly sites known for the topic.
  • Be trustworthy — linking to high-quality sources to back up claims, and have backlinks referencing it for the topics it focuses on.
  • Be updated regularly — keeping content fresh and accurate.
  • Be relevant to search intent — it should provide the information that the user is searching for, rather than a keyword-stuffed sales letter masquerading as a blog.
  • Be easy to read and get information fromthat means no massive walls of text, which ChatGPT tends to create.

Some of these are also indirect ranking factors: when your content meets these standards, your readers are more likely to read it, share it, and link to it — thus improving its ranking.

Contrary to previous statements, Google has stated that it doesn’t intend to penalize sites who use AI to write content, as long as the content retains high quality standards and is useful for humans. 

twitter screenshot - google policy on the use of AI content

So what SEO mistakes might you be making with ChatGPT that sacrifice your content’s quality, helpfulness, trustworthiness, or authoritativeness?

7 Common Mistakes When Using ChatGPT for SEO

Mistake #1: Letting ChatGPT replace firsthand experience and knowledge

Aside from all the misinformation problems we’ve all seen or read about, ChatGPT still struggles to write at the level of an expert for most topics. This damages your SEO for a few reasons, such as:

  • Lack of originality: ChatGPT uses text it scraped from the web to craft its responses (blogs, tweets, forum comments, etc), and so it requires more human input in order to produce original content. Asking it to simply “write about X” isn’t going to produce anything particularly worth reading. 
  • Lack of usefulness: ChatGPT tends to generate content that sounds overly simplistic or vague when attempting to explain complex technical concepts, making it difficult for your audience to understand the material.
  • Lack of accuracy: When ChatGPT tries to incorporate highly technical terms and concepts, it often uses them in a way that is confusing or incorrect, which can damage the credibility and authority of your content.

Remember, the most successful content for SEO needs to be original, useful (and/or entertaining), and accurate/authoritative.

Solution 

Use your own text and notes to feed into the chat – background research, quotes, notes, first-hand experience, etc. ChatGPT draws heavily from the text and key phrases in each conversation (often even if you ask it not to). In this sense, you’re providing it with extra input and context to draw from.

Bland prompts get bland results. High quality writing with unique vocabulary, a depth of knowledge, and well-written syntax will give you a much better response.

Mistake #2: Using ChatGPT for keyword research

Let’s get this out of the way: ChatGPT is not a standalone keyword tool.

I’ve seen countless people across LinkedIn, Twitter and YouTube suggesting you use ChatGPT to find long-tail keywords to optimize for (i.e. keywords with specific search intent and lower competition in search results).

Here’s how SEMRush illustrates long-tail keywords:

 

SEMRush graph explaining long tail keywords

First off, ChatGPT doesn’t have access to search volume metrics (though at the rate things are going, I expect marketing companies like SEMRush or Moz will roll out tools or plugins for that soon enough).

In other words, the results you get might be real phrases someone might use, but if you plug it into your SEO tool, the keyword isn’t particularly valuable.

For instance, what happens when you ask it for some long-tail keywords?

ChatGPT screenshot showing how someone might use ChatGPT for keyword research

Some of these keywords seem promising, but almost all have zero search volume. Only #2 is getting any traffic (“How to measure content marketing success” has 50 searches per month, and the keyword difficulty is 60). This keyword has top-of-funnel (informational) search intent, so it hardly seems worth your time to prioritize unless your site already ranks high for related keywords.

As far as actually ranking in search results for any of these zero-demand keywords? You’d still need the search results to be relatively uncompetitive, and your article would need sufficient backlinks (meaning your content needs to be good enough for people to link to it in the first place).

If you simply optimize for keywords based on a list it provides you, you’re going to be disappointed with the results.

Solution

For now, ChatGPT is a starting point for keyword research, such as asking for LSI keywords to help you brainstorm ideas. From there, you can use an SEO tool like the “related” keywords option on SEMrush to expand your search.

ChatGPT screenshot showing how to find LSI keywords for SEO

Mistake #3: Failing to fact check AI-generated text before publishing 

If you’ve read even half a headline about ChatGPT this year, you know one of its biggest issues is its tendency to “hallucinate,” or in layman’s terms, “confidently spew misinformation.” 

Even if you’re an expert in the topic you’re writing about, it’s easy to overlook these false claims while reviewing ChatGPT’s responses.

ChatGPT screenshot showing a conversation about claims accuracy

A lawyer I spoke with for this article said he tried using ChatGPT to provide legal briefs related to a case he was working on. ChatGPT provided a list of cases and outcomes, with correct formatting, and convincing legal language. At first glance, it seemed encouraging.

Yet, when he cross-referenced these cases with his firm’s legal database, he found ChatGPT’s responses either had (1) incorrect dates, (2) incorrect outcomes, (3) were completely fabricated, or (4) referenced a real case, but gave incorrect details about it.

So what happens when you use it to write an article that completely neglects new trends in your audience’s industry, or gets key facts wrong that anyone with sufficient industry knowledge would recognize immediately? 

It kills your credibility, and damages trust with your audience. That’s a nightmare for a successful marketing funnel, and it hurts your search traffic in the long run.

Solution

It might seem promising to ask ChatGPT for URLs or citations, but these also tend to be fabricated. A better approach is to load your vetted research into the chat, and use that as a starting point for ChatGPT to draw data from.

Tools are also popping up to make this easier: Bing’s first rollout of its ChatGPT integration has had mixed results, though other tools like Perplexity and Scite are working on combining AI writing capabilities with better citation.

Perplexity.AI screenshot about improving SEO, shows citations

Mistake #4: Failing to revise AI-generated text for quality

In a few years, an AI tool like ChatGPT will probably sweep the floor with today’s content writers, but right now the only sweeping it’s doing is when making generalizations.

In other words, copy+pasting ChatGPT’s responses into your posts verbatim is a recipe for derivative content that readers (and therefore Google) will ignore. 

Even after accounting for misinformation, ChatGPT leaves a lot for editors to fix:

  • Verbose phrasing: AI frequently relies on fluff phrases like “In this blog, we will discuss…” or “There are many ways that….” or  “In conclusion…” to fill up space rather than add useful knowledge.
  • Repetitiveness: Longer responses tend to get repetitive. Pick a key adjective or term like “powerful” or “efficient” in its response, and check e how many times it shows up. I’ll bet you it’s a lot.
  • Generic advice: How-to content in ChatGPT tends to produce underwhelming suggestions. For instance, anything it tells you about marketing will often center around social media. Without help, it struggles to produce original content that would rank higher in search.

ChatGPT screenshot showing how ChatGPT can give clearly false results that sound credible

Take this screenshot for instance: ignoring the fact that ChatGPT thinks it’s known for live videos and allowing brands “to interact with followers in real-time,” this intro focuses entirely on social media, and takes about 3 times longer than necessary to make its key points — which aren’t exactly correct in the first place.

Solution

When it comes to sentence quality, your fastest route is usually to rewrite text yourself (starting by deleting all the excess fluff). GPT-4 improves this issue to a certain extent, but it’s not perfect either. You can also give follow-up prompts, asking it to elaborate or rephrase the response. The more specific you are with your request, the better your results will typically be. 

This reply prompt I gave it below is pretty basic, but already gives me much better results than my first prompt, and would require less revision before using:

ChatGPT screenshot showing how to improve content quality by replying to original response

Mistake #5: Having ChatGPT write a full article all at once

It’s tempting to save time and churn out all your text in one go, but the longer the final text, the more guidance ChatGPT tends to need.

Otherwise, it starts to recycle phrases really quickly, making the same points and losing any sense of overarching narrative.

Solution

Using your original research or notes as a starting point, have it make an outline, then edit the outline to your needs and expand it in ChatGPT, section-by-section, to get closer to something approaching a first draft. More specific requirements in your prompt will generally yield better results.

Mistake #6: Using ChatGPT to optimize content for a keyword

Let’s assume you’ve found some great keywords to focus on. Now you just have to tell ChatGPT to write some content optimizing for that keyword, right? Well, not quite.

ChatGPT doesn’t always have the best grasp of a keyword’s meaning, and often fails to weave it naturally into text. It also has a tendency to overuse keywords when you ask it to optimize for one. While keywords are important for SEO, overusing them can result in content that feels forced and robotic, and panders to search engines rather than readers.

Plus, Google’s web crawlers will classify your content as lower quality if you do this. 

While yes, admittedly, trying to “game” search engines like this can sometimes work in the short term, it’s usually just one Google update away from being sent to the dark recesses of the internet, never to be clicked on again.

Solution

In your prompts, you CAN try specifying that ChatGPT weaves in a keyword organically X number of times, or asking it to write variations of the keyword. This can work, so long as that’s not the only requirement you provide in your prompt.

However, the real trick here is to focus on producing quality, comprehensive content for the scope of your article — when you do this, the most relevant words and phrases should come up naturally. As mentioned above, feeding your own notes to ChatGPT goes a long way.

As Google algorithms become more sophisticated, they put less weight on the individual keywords, and more on the content of the sentences and paragraphs as a whole. In other words, an informative, original take on your topic will perform better than a generic piece of content stuffed with keywords.

Mistake #7: Writing about current topics with ChatGPT

As of this writing, ChatGPT draws from texts published in 2021 or earlier. In other words, current events and breaking news are going to be missing, or dangerously incomplete. 

ChatGPT screenshot showing how commonly known facts after 2021 can return incorrect results

The state of the economy, the latest updates to you or your competitor’s product, the most common problems of your target customer… if you try to rely on chat responses, you risk basing your content on outdated information, and coming across to readers as out-of-touch.

Warning: You might be thinking “well I can just paste a URL into the chat and have it summarize that information,” but you’d be out of luck without the right plugin or app. ChatGPT alone won’t actually read the content within a URL: it just makes a very educated guess about what’s on the page, based on the domain name and keywords in the URL slug. You can test this by making up a URL and seeing what ChatGPT says about it.

ChatGPT screenshot showing how asking it to summarize a URL doesn't work

Solution

  • When writing about any current topic, use the tip we described above and feed reference material into the conversation, such as subject matter expert interviews, research papers, and other high-quality writing, in order to give ChatGPT access to current information.
  • If you’re using ChatGPT to talk about your brand, make sure to give it your most up-to-date offers, messaging, and positioning at the start of the conversation. This might take longer on ChatGPT than it would to write it all yourself, but if you plan to regularly write about your brand, you can keep this information on hand.

What should I use ChatGPT for?

ChatGPT can help you write much more content in less time, but it’s still no replacement for the human judgment necessary to properly optimize your content for search, or to create valuable content as part of a cohesive content marketing strategy. 

It’s a powerful tool — but for now, still a tool.

But there are a few things ChatGPT is good for. For starters:

  • Inspiration — helping you come up with titles, headlines, examples, etc. Ask it for a list and use the best result as a starting point
  • Summarizing — paste in your raw text and have ChatGPT summarize it for you, or explain it in simpler language
  • Rephrasing — if you’re a subject matter expert, but not a writer, you can use ChatGPT to rewrite your thoughts more clearly

Check out our post on how to use ChatGPT for marketing to see how else you can use it.

“Original content” is purposely not included here. You’ll need to supply your own data, opinions, and editorial decisions to generate content that drives results — whether in your SEO or in your overall content marketing strategy.

Original expert content is the foundation of successful SEO and content marketing. Oftentimes, you might already have great expert content such as a video or podcast, but you don’t have the time to properly repurpose it into a blog or other content asset.  If you’re looking for ways to multiply your content production without sacrificing quality, consider a content repurposing service.

Next Post

Oracle Stock Nears 2021 Highs; Big Tech Firm Teams Up With Nvidia On AI

Big Tech Oracle (ORCL) has been on a tear since its March earnings report. The Big Cap 20 company teamed up with an industry powerhouse in its AI initiatives. X The Austin, Texas-based behemoth has a $274 billion market cap, having grown into one of the largest players in cloud […]
Oracle Stock Nears 2021 Highs; Big Tech Firm Teams Up With Nvidia On AI