How to Use Copysmith to Write Articles That Actually Rank

AI content generation is a hot topic in marketing today. As the tools evolve and grow more sophisticated, and as more and more businesses white labeling them pop up, they seem to be everywhere. Some people hate it, some people have lost their jobs because of it, and others have gotten rich off the backs of the tools.

One such tool is Copysmith. It can be powerful – and big brands like Pfizer, Ford, and Dell use it – but like all AI tools, you have to use it carefully if you want to produce anything worthwhile.

The internet right now is a churning pool of AI sludge just beneath the surface, and it takes some serious effort to rise above the rest. Where do you guide that effort?

How can you use an AI tool to produce content that isn’t just more sludge?

I have some tips to help.

What is Copysmith and How Does it Work?

Before we get into how to use Copysmith, I need to talk about what Copysmith actually is for two reasons.

The first is just as a general profile of the service, and the second is because there’s some very confusing nonsense happening in Google’s search results that I want to bring up.

Addressing the confusion.

First, let’s talk a little about what Copysmith is and how it works. I want to specify this because there are actually several “Copysmiths” out there.

• Copysmith.ai, which is the one we’re talking about.
• AI-Copysmith, which is a shell laid over GPT-3.
• Copysmith.ie, a Dublin-based digitization service that probably isn’t happy its name has been effectively overridden by AI services.
• CopySmiths, an Australian content marketing agency.

It’s not exactly an uncommon name. After all, a smith is a creator, copy is content, so “copysmith” is a low-hanging fruit of a name.

A big part of why I wanted to specify here is that copysmith.ai and ai-copysmith.com are very confusing. It seems like ai-copysmith is an older version of copysmith.ai – they even have the same logo, essentially, though one is colored and one isn’t – and there’s nothing on either site tying the two together. It looks like two different services at first glance.

It doesn’t help that the search results are mixed too. Just look at this screenshot from Google:

The image pack at the top is a muddled mix; some of the screenshots are for ai-copysmith, and some of them are for copysmith.ai. The text result on the left is for copysmith.ai, but the descriptive text on the right is for copysmith.ie, with an incorrect founding date and subsidiaries.

Then there’s the simple fact that if you go to ai-copysmith.com and click to start a free trial, it doesn’t work. They used ClickMagick for referral tracking, and their account has been canceled. So, I guess it’s just a dead service that gets to rank #2 on Google now?

So I don’t blame anyone for being confused here.

Here’s what I believe happened, after some investigative research. I think ai-copysmith.com was the business Copysmith, acting as an interface for ChatGPT-3. When that started to fail, they sought seed money and used it to buy better services, namely Frase, Rytr, and Describely.

They then retooled their site into Copysmith.ai, made those services their main product, and essentially abandoned the old site. Meanwhile, Google is trying to sort through all of these and showing us some of the other companies that share the same name.

Analyzing the actual Copysmith.

For the purposes of this post, I’m solely talking about Copysmith.ai. You know, the one that works.

As I just mentioned, Copysmith is basically a trench coat with three services hiding in it.

Describely is a short-form generator meant specifically to create product descriptions, product detail boxes, and catalog entries. It’s effectively a way to automate the drudge work of filling out a product catalog. In fact, it’s one of the few uses for AI content generation that seems reasonable even to many anti-AI folks; it’s barely more than a MadLib keyword inserter, but it’s a way to automate a lot of tedious labor that doesn’t really require skill but does require time.

Rytr is a longer-form AI generator with the twist that it’s meant to be trained specifically on your unique voice. You feed in a bunch of your own content, and it analyzes your voice, style, tone, and word choices and mimics them when you have it generate things like emails, ad copy, and shorter blog content.

Frase is a long-form content generator focused on taking a set of keywords and generating full-length content. It has a bunch of more marketing-focused data and tools on the side, like a topic score analysis, keyword counts and density analysis, and so on. You can use it for basic research, outlining, and content creation.

I’ve mentioned both Frase and Rytr before here on Topicfinder, and while I’m not all that impressed with either one, the goal of today’s post is to leverage them to create something worthwhile, not to criticize them.

So, how do you actually use these tools to create something that can rank in Google’s search? Other than by simply generating 10,000 blog posts a week and hoping the shotgun method works, of course. Pro tip: it doesn’t.

How to Generate Content That Actually Ranks

Despite going into the specific tools in the first chunk of this post, I’m not actually going to give you a tutorial on using Copysmith/Frase/Rytr/Describely. That’s because everything I have to offer below is applicable to every AI content generator, because at the back end, almost all of them are using the same small handful of LLMs, so they all have the same problems.

So, whether you use Copysmith, go to ChatGPT directly, or use Claude or Jasper, or whatever else, these steps apply.

Pick a subject that isn’t doomed from the start.

The first thing you need to do is pick a focused topic that actually has a chance to rank at all.

The biggest issue I see people who rely on AI running into is this initial hurdle. It’s not a hurdle you can surmount with AI tools, either, because of how LLMs work.

LLMs function by being trained on massive amounts of content and then using very complex math to produce content that is linguistically similar to the content that already exists. It is, by definition, going to produce things that are close to what already exists. Creating anything unique goes against the core design of the systems and how they work.

That’s why a huge amount of AI prompt creation is all about layering on dozens or hundreds of instructions meant to wrench the generator away from the most common end result. It’s also why immense amounts of “AI-generated” content are word-for-word identical to pre-existing content.

All of this means that if you just plug in a keyword and ask an AI tool to generate content, you’re going to get something that substantially resembles – if not outright copies – the content in the top few Google search results for that keyword.

The topics generated by the AI have a bunch of problems immediately:

• Topics that are too competitive for your site. If you put in a keyword like “what is SEO,” do you think you’re going to out-rank Google itself, Moz, Wikipedia, Search Engine Land, and the other big names already up there? Certainly not.
• Topics that have essentially no traffic. AI isn’t designed to have any idea what a fact is or to think at all, no matter how much people like Sam Altman try to convince you that AGI is real. Its only goal is to output a sequence of words that resembles human language. It can very easily be nonsense that looks fine at first glance. Then you check, and you see that it’s not a query anyone would ever care about, like “What are the benefits of chocolate-flavored motor oil?”
• Topics that don’t match user search intent to content focus. One of the most important parts of modern content marketing is taking a topic and figuring out not just what to write about but what the people searching for that topic are hoping to find. You need to understand search intent and write (or generate) content that matches that intent.
• Topics that are saturated with enhanced results. Another common issue is that topics that might seem good can actually be awful because there’s a near-zero chance that anyone searching for them wants to look at a blog post. A keyword like “Chinese restaurants near me” is going to be dominated by local business results and the map pack; your blog post might be a great list of reviews (though if you’re relying on AI to make it, it probably won’t be), but it’s not going to show up at all.

Any AI tool claiming it can help produce this kind of topic for you is probably only barely using AI and is instead using more reasonable topic and keyword research tools with actual data behind them and just presenting it in a way that looks like AI created it for you. Meanwhile, you can just do the research yourself and apply a little critical thinking for much better results.

Make sure your title is excellent.

Another very common issue – which also stems from the same LLM problem of seeking similarity to existing content – is that AI-generated titles just aren’t very good.

AI-generated titles almost always fall into at least one of these categories.

• Extremely basic, like “The Top 10 Products in Category.”
• Overly verbose, like “Exploring the Nuances of Using Products in Category.”
• Identical to the top search result.
• Identical to other people’s AI-generated content and titles.
• Too long and won’t fit in Google Search because they exceed the pixel limit.
• They always start with “ing” words, aka Gerunds (e.g. “Understanding The…”) or words like “Discover”, “Delve”, etc.
• They are too concise and remove many “optional” words (e.g. the, that, this, which, to, etc).

Not only are they not interesting, but they’re also easy to spot, especially from a top-down perspective like Google’s, where they can see your whole blog and analyze all of your titles at once.

This is where I toot my own horn a bit. To solve this problem, all you need to do is use Topicfinder. It gives you a bird’s-eye view of all of the titles and topics of competing content and gives you great ideas for your own versions and variations.

Be as specific with your prompting as possible.

There is no world where you can go to an AI tool and type in “write me a blog post about SEO” and have it output anything that will ever rank for anything with any competition or value. You have to put in more work than that.

Prompting is usually a multi-stage affair these days, which is why most AI content creation tools also break things down into individual questions they ask you to answer to fill out a prompt on the back end for that paragraph or section. You’ll end up needing to prompt for a basic outline, revise and review the outline, prompt for each section, revise and review each section, further prompt to optimize those sections, and put it all together.

You can sometimes get away with refining a lengthy prompt template and combining multiple stages, but you run the risk of something going off the rails in an early stage and needing to fix it anyway. I find an iterative approach is most effective.

Truthfully, sometimes writing the prompt ends up taking as long as it would have taken to write the post in the first place, but the goal is repetition, not single-instance creation, so it can still be better over time.

Review and edit the content.

If you take any piece of advice from this entire post, let it be this one. AI content is always in need of editing to add the human touch to it. Your goal is to edit and revise your content in a variety of ways, aimed both at “fooling” the AI detectors and at improving the content.

AI makes mistakes, AI hallucinates, and AI can generate absolute nonsense. Your job is to know enough about your subject to avoid those problems.

• Ensure the tone, POV, and voice are consistent. This is less of a problem these days than it used to be, and your prompts will generally include tone and voice instructions, but you still need to review it because the AI doesn’t actually know what a tone is or what a POV is, just what words are tagged with those identifiers in the training data.
• Relatedly, make sure the tone, POV, and voice are consistent across blog posts. Unless you have multiple “authors” on your site, ostensibly, everything is written by the same person and should have the same voice.
• Add your expertise in the subject. At some level, your content needs to be unique and valuable, and the AI can’t do that for you without hallucinations. You need to be able to add your own expertise.
• Check for inaccuracies. Hallucinations are a huge problem and one of the biggest tells of AI content. They also wreck the credibility of a blog and can hurt your whole site. Fact-check your posts!
• Add personal anecdotes. Having personal experience to relate to always helps. I find that a few first-person interjections also help liven up blog posts and make them feel more casual for readers to enjoy.

Most of this is reflected in EEAT, the Expertise, Experience, Authoritativeness, and Trustworthiness metrics Google has been pushing for the last few years.

Add internal and external links.

One thing the AI can’t usually do is add links. Fortunately, adding source links isn’t difficult, though it is time-consuming.

Just go through your posts and look for chances where you can add internal and external links. There are also ways to add internal links automatically, though I don’t recommend it for external links.

Add images and other media (that aren’t generated).

Well-done AI-generated content can slip under the radar, and if it’s fact-checked, it doesn’t hugely matter if it was substantially generated rather than written as far as the audience is concerned. AI-generated images, on the other hand, are a huge tell, a huge loss of credibility, and a significant hit to the value of your site. Despite recent advances, most AI image generation is still grody.

On top of that, courts have recently reaffirmed that AI images can’t be copyrighted, which can be an issue down the line.

So, enrich your posts with images and other media that add value and don’t drag you down.

Only when all of this is done will you have blog posts that have a chance to rank. Even then, you have all of the usual challenges – competition, promotion, technical SEO, etc. – to confront. You just have to do all of the above to have a place on the playing field.

Leave a Comment