Users can use generative AI to create and edit content directly from within Marketpath CMS. All generated content comes from third-party service providers, however, and the quality of the content - as well as the cost to generate it - is dependent on the model of the service provider. For most purposes, the default provider and model will provide the best balance of cost and quality. However, if desired, the user may choose to select a different model from the ones supported by Marketpath. Read on to learn more about how generative AI works in Marketpath CMS and the options available to you.
To generate new content in the HTML Editor, click the AI toolbar button and the Generate Content Using AI dialog will open. It will automatically suggest a few ideas for you, such as "Write a blog post" and "Insert filler text". If you already have content in the HTML editor, it will also display a button to "Edit my existing content". Feel free to select one of these ideas and follow the next prompt, or to write your own prompt below. When you are satisfied with your prompt, click on the send button to submit your prompt to the AI service.
When the new content is generated, it will be displayed at the top of the dialog, and you will have two options:
The original content of your HTML editor will not be changed until you click the Use Text toolbar button. At that point it will replace the text you had originally selected with the generated content (or all text if you clicked on the "Edit my existing content" button).
Additional options in the dialog:
You can also chat with AI directly from Marketpath CMS. AI Chats are associated with a specific site, and utilize the AI tokens of that site. Other users can also view and contribute to the chat. You can also view and continue your own chats on other sites using the AI sidebar.
To start an AI chat in Marketpath CMS, click on the AI sidebar tab and enter your prompt at the bottom of the sidebar. There are a few options you can set before you send your prompt, including selecting the provider and model you want to use, whether you want to generate a text or html response, and what to name your chat. When you are ready to submit your prompt, click the go button.
Once the chat has been started, simply enter a new prompt to continue chatting with AI. You can change the provider and model as well as the response type any time you want - even in the middle of your chat.
To view your chat history, click on the Chat History button in the top right of the sidebar. By default this will list your chats on this site, but you can quickly change it to list all chats on this site or your chats on all sites instead. From here you can continue, rename, or delete chats, although you can only rename and delete chats that you started. Also note that deleting chats does not negate the AI tokens used by the chat.
AI tokens represent the cost to generate content in Marketpath CMS. Every AI prompt requires one or more AI tokens. The exact number of tokens for each prompt depends on the model used, as well as the complexity of the prompt and the length of the content.
AI tokens are an important abstraction because they allow Marketpath to use the same pricing system for different providers, models, and types of content. They also make it possible to adjust our cost calculations over time while keeping customer costs predictable, even when the products and pricing from third-party providers change.
Every new and AI-enabled site has a specific number of AI tokens available. However, active sites that do not have an AI-enabled plan will be required to upgrade to an AI-enabled plan before they can use any AI tokens. Refer to the pricing page for specific details about which plans have AI enabled, how many tokens are included, and what the overages cost.
To summarize, there are three types of sites, each handling AI differently, as follows:
New sites receive a small number of free AI tokens during initial development, regardless of which plan they have selected. These tokens are only valid before the site is activated, and it is not possible to get more AI tokens without first activating the site. Once the site is activated, any remaining free tokens are removed and replaced with the number of tokens specified by the site’s plan.
Each AI-enabled plan includes a pre-specified number of free AI tokens each month. Unused tokens do not carry over to the next month. If you use all of your free AI tokens in a month, you have three options:
Once a user allows overages for AI tokens on a site, any user with editing access to the site may continue to use AI tokens on that site until the end of the current billing period. At the end of the billing period, the overage cost will be calculated based on the total number of AI tokens used that month and the overage costs specified in the plan.
Sites will not be charged overages unless an authorized user accepts the billing overages for that month. Once the next billing period starts, a user must accept the billing overages again to utilize more tokens than are included for free in the site’s plan.
Sites that have been activated but do not have an AI-enabled plan will not be able to use any AI tokens and will not be able to generate content with AI unless the site is upgraded to an AI-enabled plan.
OpenAI is the biggest artificial intelligence company and has been on the leading edge of AI research and services since generative AI came to public awareness. They received massive public attention and sparked the AI race with the release of their free AI chatbot - ChatGPT - in 2022.
Links: About OpenAI, OpenAI Models, OpenAI Privacy Policy
This is the default model, which generally provides good results at a cost-effective rate. Every prompt sent to the GPT-4o Mini model costs one token regardless of the prompt and response lengths or whether or not it is part of a chat.
This is the more powerful version of the GPT-4o Mini model. It is capable of generating more complicated and higher-quality responses to your prompts, but that comes at a cost. Prompts sent to the GPT-4o model cost 1 AI token for every 3000 input tokens plus 1 AI token for every 1000 output tokens, with a minimum of 2 tokens. According to the OpenAI documentation, 1000 input/output tokens is approximately 750 words.
When sending the prompt as part of a chat, the entire chat history is sent along with the prompt, which may increase the cost as the chat grows longer.
These providers and their models are not currently available, but could be added in the future. Please ask if you would like to see any of these in Marketpath CMS!
Google is the provider behind the Gemini AI models. Marketpath does not currently have an integration with Google or Gemini models, but please ask if you would like to see us add this in the future!
As stated on their website, Claude is a family of state-of-the-art language models developed by Anthropic. Marketpath does not currently have an integration with Anthropic or Claude models, but please ask if you would like to see us add this in the future!
In addition to their own Llama models, Meta (a Facebook company) has implementations of numerous other AI models available through their API. Marketpath does not currently have an integration with Meta or Llama models, but please ask if you would like to see us add this in the future!
The stated goal of AI21 Labs is to build AI systems that become thought parterns for humans. Their flagship Jamba Models are built for speed and long context windows (so you can send longer input). Marketpath does not currently have an integration with AI21 Labs or Jamba models but please ask if you would like to see us add this in the future!
After the public release of their R1 model, Marketpath attempted to obtain an API key to test their API and make it available to our customers who also wanted to try it out. Regretfully, their platform was "offline for maintenance" for so long (a week at the time of this writing) that we gave up trying to get an API key. So while we now have the technical ability to send prompts to the DeepSeek models, we will have to wait until they are more stable to make it available to our customers.
DeepSeek is a Chinese AI company that made a splash in the AI scene by training a high-quality AI model that rivals the leading models from other providers but at a fraction of the cost. Their stunning and inexplicable success has caused speculation regarding the veracity of their claims and investigations into their business and other practices.
Links: DeepSeek Website, DeepSeek Models
A cost-effective content generation model, generally considered comparable in quality to OpenAI's GPT-4o model. Every prompt sent tot he DeepSeek Chat model costs one token regardless of the prompt and response lengths or whether or not it is part of a chat.
The flagship model that earned DeepSeek global attention. This model is generally considered comparable in quality to OpenAI's o1 model (not currently supported by Marketpath). Prompts sent to the DeepSeek Reasoner model cost 1 AI token for every 10,000 input tokens plus 1 AI token for every 1000 output tokens. According to the DeepSeek documentation, 1 english characcter is approximately 0.3 input/output tokens.
When sending the prompt as part of a chat, the entire chat history is sent along with the prompt, which may increase the cost as the chat grows longer.
Generative AI refers to technology that can create new content based on patterns and examples from existing data. It can answer questions, create and edit content, and perform simple actions.
An AI model is a mathematical representation or algorithm trained to perform tasks like creating new content based on previously analyzed data. Models are trained on large datasets, and their capabilities are limited by the data available at the time of training.
An AI Provider is a company that creates, trains, and maintains AI models for public use, each handling data and programming differently.
AI Tokens are a system used to track the cost of generating content. More complex prompts cost more tokens, and usage depends on the specific AI model and request length. Fore more information about the cost of various models, read Providers and Models.
We are excited to be able to generate images in the near future, but this functionality is not complete yet. Keep your eyes open for this soon!
You have full rights to the AI-generated content and can use, modify, or ignore it as desired.
AI is included in Plus and higher-tier plans and as an add-on to the Standard plan for an additional $10/month.
Token availability varies based on your plan and resets monthly. You can use AI until your tokens run out each month, or if you accept the overages. For example, 400 tokens per month is the equivalent of 400 prompts per month if each prompt is 1 token, or 100 prompts per month if the average prompt is 4 tokens (based on the model selected). You can read more about the token system in Tokens and Pricing.
You can use or modify the generated content freely, but it is advisable to enrich it with your own knowledge to avoid detection and penalization by search engines.
The accuracy of AI data should be verified as it may not always be accurate or up-to-date. Excess AI-generated content without human enrichment can attract search engine penalties. In some cases, AI-generated images may have legal risks concerning copyright - particularly when copying or modifying images from other sources or creating images in a particular artist's style.
Your needs determine the best AI model. Some models may be more up-to-date with better results, while others prioritize faster processing. More information about AI models is available under Providers and Models.
Anyone with editing access to your site can use AI if it is enabled on the site. Consider restricting access if AI token abuse is a concern. Create a support request if you would like to further limit access to AI on your site.
Currently, you can generate text and HTML content. Support for AI-generated images is coming soon.
Yes, you can use externally generated AI content from providers like OpenAI or Gemini. Our CMS integration just simplifies this process.