This was originally published in our monthly newsletter. Subscribe here.

I’ve been kicking the AI tires to produce product messaging. Maybe we’re all ChatGPT’ed out, but I haven’t seen anyone do a side-by-side comparison with other tools and I have a real-world use case in the works.

I‘m comparing ChatGPT3, Jasper, and Copy.ai for creating some basic product messaging:

Product Description
Positioning Statement
Pitch to the Buyer
Value Propositions

I’m using Padex as an example, an early-stage patent analyzer tool. Full disclosure, Padex is built by my husband, Ryan, and I’m helping him take it to market (but it hasn’t properly launched yet). I know I’m talking my book, but here’s why:

I wanted a company in an uncharted category to see how generative these AI tools actually are.
I couldn't think of a cool fake example.
I’m already doing this work in real life.

I’ve documented the process — what my prompts were, how long it took, money spent, and what I think is the best way to use these tools.

Product Description

Existing Messaging: AI patent analyzer to review PDFs faster — Padex saves time and improves quality for patent analyzers using AI to connect every component of your patent. Used by folks in Intellectual Property, Freedom to Operate, and Patent Litigation.

Positioning Statement

Existing Messaging: People who work in Freedom to Operate and Patent Research who want to save time, reduce friction, and improve accuracy when analyzing patents choose Padex because it performs 2X better than Google Patents.

💡 Simple positioning statement formula: [Target market] who [need or want] [unique value proposition] choose [brand] because [reason to believe].

Positioning Statements for Padex generated with AI writing tools

Pitch to the Buyer

Existing Messaging: Padex uses AI to automatically recognize and connect components of your document so you can review patents faster. No more scrolling back and forth or keeping several windows open at once. Padex retrieves 2X more information than Google Patents. Sign up to join the beta for free.

Value Propositions

Existing Messaging:

Saves time searching and analyzing patents by making it easy to find and jump between related figures and text.
Less frustration and toil while investigating patents; no more having multiple windows open and scrolling back and forth.
Improved accuracy to actually find all the components of the patent — nearly 2x more accurate than Google Patents.

So how did they stack up?

I got the most benefit from ChatGPT and Jasper. I even started on a blog post, playing ChatGPT and Jasper responses off of each other (using concepts discovered in one to paste as prompts in the other). You can see screenshots of my transcripts from each tool in this folder.

ChatGPT — produced the most flexible, generative, and long-form content

Spent: 2 hours on this project and purchased $18 in credits, used $0.01

I enjoyed using ChatGPT the most, and it was the most flexible. It was able to generate large amounts of text (a big limitation of Copy.ai), and I quickly started using it as my personal brainstorming buddy who never has to go to another meeting or judges me for crappy ideas. Despite scaling issues, it’s quite fast, but I recommend copying/pasting results as you go so you don’t lose them.

ChatGPT was able to answer sidebar questions in the middle of chat sessions. For example, after I asked for a positioning statement for Padex, I asked for a positioning statement formula, which it provided. I then used the formula in the next prompt.

I know I’ve only scratched the surface, and becoming good at prompt engineering will take time. I will continue to explore the opportunities!

Use ChatGPT to:

Brainstorm broad topics with a tighter feedback loop than a Google search
Generate content iteratively (i.e., give it feedback and get more specific as you go)

Jasper: Chat & Templates — most structured and immediately useful

Spent: 1.5 hours on this exercise; purchased a $590 annual plan back in September

You should know Jasper is built on OpenAI, but is essentially forked for marketing use cases.

My first asks on the product description were *lightning* fast. But I was underwhelmed by the answers compared to ChatGPT. Some answers feel like it’s just shuffling copy around rather than giving new content or ideas.

That said, I appreciated the structured templates (think of them as premade prompts), and the templates themselves generated ideas. Jasper had great structure, speed, and output with a smaller learning curve than ChatGPT.

Use Jasper to:

Iterate quickly on existing messaging/copy
Produce baseline blog content and social posts
Optimize existing content for SEO

Copy.AI — feels like a copy slot machine, lacking templates and prompts

Spent: 30 minutes on the free plan; wasn’t compelled to upgrade

I was least impressed with Copy.ai. If you already have an idea of what you want to say, I think it’s great for cranking out ad copy and headline alternatives, but it’s not generating much, and the form of the content is limited. It has fewer templates and less prompt flexibility, which meant some messaging couldn’t be created directly (like positioning statements and the pitch to the buyer).

With generative AI we will shift from creators to narrators

There’s a lot of fear that humans will be replaced by AI (I’ve noticed this feeling is especially strong among content & social marketers). This may happen in the long term, but in the short term, we will use these tools just like every other kind of automation: to get more done and spend more time on leisure. When the car was invented, we didn’t just go to the same nearby locations faster, we went further than we ever imagined.

David Friedberg’s core thesis outlines this best, which I’ll paraphrase:

“As we started to do work on earth and engineer the world around us, we did labor to do that. Over time we built machines to automate our labor. From tractors to microwaves—we became less dependent on our labor. And we got to switch our time to knowledge work, as creators. The vast majority of the world now spends their time on knowledge work, supported by computers. As the software and AI supplants knowledge work, the role of humans transitions to the narrator. Instead of filming a movie, you will narrate the movie you want to see and the software will create it for you. Instead of drawing blueprints for a house, you will dictate the house you want to build. Instead of creating the content, you will narrate the content you want to experience.”

— David Friedberg on All-In Podcast Ep 99
(his take is inspiring, I recommend giving it a listen)

Things to keep in mind:

Selection bias means only extremely good and extremely bad results are trending on social media.
These tools don’t know what they don’t know. Ask ChatGPT “how many bears has Russia sent to space?” and it answers: “Russia has sent two bears into space. The first was in 1983 and the second in 1985. Both bears were named Belka and Strelka and were part of the Soviet Union's Kosmos 18 mission”. ChatGPT sounds sure of itself and the answer is grammatically correct, but it doesn’t have a shred of common sense.
Changes are likely coming for Google’s search algorithm to deal with the sheer volume of content that will be produced. Perhaps last-touch attribution will be weighted even more heavily (rather than quantity, frequency, or recency). Personally, I’ll be focussing on getting very high-quality content out the door sooner, not on creating more mediocre content.
Yes, there are legal/copyright issues looming, but the benefits of AI tools are too good to avoid for fear of potential legal implications. I’m sure guardrails will be introduced and we’ll make adjustments, but there is plenty to be gained in the meantime.

Comparing AI-writing tools to create real product messaging