On the one hand, there are plenty of simple and free alternatives for generating a single product description with GPT. On the other hand, custom CMS integrations generate them in bulk but can be pricey and/or leave you with little control of the input and output parameters.
On this page, I’ll explore a couple of middle-ground options that may be suitable for small and medium-sized retailers with somewhat large product catalogs.
Free Vs. Paid Product Description Generators
The number of AI applications is growing exponentially and the alternatives vary widely in terms of functionality. At the time of writing, OpenAI’s LLM ChatGPT in version 3.5 (GPT-3.5) is the most common in both free and paid tools, although some also support the considerably more resource-intensive GPT-4 model.
Free AI Description Generators
Either of these models can be used free of charge via OpenAI (GPT-3.5) or Microsoft’s Bing Copilot (GPT-4), respectively, among other places. There are also free options that cater directly to e-com retailers such as the product description generator tool from Ahrefs.
However, generating a large number of product descriptions in that way is not very practical. Generating descriptions in bulk is obviously far more efficient, and the following are a couple of examples that can do just that without breaking the bank.
Bulk Generation with Writesonic
Writesonic lets you choose between GPT-3.5 (‘Premium’) and GPT-4 (‘Superior’) in an easy-to-use interface focused on general content generation. The ‘Bulk Generate’ option is available for product descriptions, Ad copy, and a range of other content types.
This tool uses templates for each content type and the product description template is fairly simple. Only the ‘Premium’ quality type (GPT-3.5) can selected and there isn’t much in the way of fine-tuning.
It is nevertheless very easy to use and the output is about what you would expect from GPT-3.5 for this type of content.
As always, proofreading and editing are a must before publishing any sort of generative AI content, but this is also the case for fully automated integrations. This particular example is about 130 words of rather generic copy, but experimenting with the input in the product_characteristics column of the template may produce better results.
Bulk Generation with SheetGPT
SheetGPT is different from most of the alternatives as it isn’t a custom interface on top of ChatGPT. Instead, it is an add-on for Google Sheets.
The content is generated via formulas and there are quite a few options to experiment with. Among other things, you can opt to generate with GPT-4, which may produce slightly better results at the cost of a rather large increase in the consumption of tokens (i.e. cost).
A simple example that roughly corresponds to the above could look like this:
The result, based on the above prompt, is remarkably similar to the output generated by WriteSonic. This is perhaps to be expected since it uses the same GPT-3.5 model.
The main difference is that SheetGPT adds a more comprehensive layer of control, not least the ability to use any prompt. You could theoretically create different sheets for each of your product categories with custom, detailed prompts for all of them.
Although getting started with SheetGPT is somewhat more complex, it is arguably far more versatile.
It seems like a good idea not to get into large-scale batch processing right away, but rather fine-tune your prompts with individual products until the result is satisfactory.
Wrapping Up
Product description generation is one of the most compelling use cases for generative AI at this point in time. In general, this type of text does not require hands-on experience and it doesn’t necessarily matter if they are somewhat generic.
You nevertheless have to check the generated descriptions, as factual errors are still common, but you can still save a lot of time compared to manual production.
