Purple Prompts FAQ

Yes, and we even recommend adapting prompts because you want to constantly optimise your prompts to make your editorial work more efficient in the long term. 

A central component of Purple Prompts is the prompts management interface. The UI is for managing saved prompts, i.e. customise existing prompts and add new prompts. Besides, you can even edit the Prompt on the article view while applying it.

Find out more about how prompts management works in the Purple Hub.

In general, we recommend playing with prompts, trying different versions and tweaking reused prompts to get better results over time.

To help you get the best results from your prompts, we have summarised our 'Prompt Writing Best Practices' in an article.

We have even dedicated an entire webinar to this topic.

We have also collected some external articles that we think might be useful for getting a good basic understanding of how to work with prompts:

OpenAI prompt engineering best practices 

A collection of example prompts useful for journalists (in German)

General tips for talking to language models

You can use Purple Prompts to save time when creating content and automate manual tasks after article creation, using prompts stored directly in the Purple Hub. 

You can also rewrite the article using custom prompts, for example, 'Rewrite press release' or 'Apply reader engagement tactics'.

Purple Prompts uses large language models from OpenAI and Anthropic (Claude), which we call directly and make available in the Purple Hub. Your Hub can use either vendor, or both, and administrators choose which model handles which job, see AI Models and API Keys.

Purple Prompts sends your content to the API of the vendor your Hub is configured for, and both vendors state that API data is not used for training.

OpenAI states about their API usage:

“OpenAI does not use data submitted to and generated by our API to train OpenAI models or improve OpenAI’s service offering”

“Data is stored on the OpenAI site for up to 30 days, but only to recognise any problematic use, and then deleted again.”

Learn more on the external Website: How your data is used to improve model performance | OpenAI Help Center

Anthropic states about their API usage:

“By default, we will not use your inputs or outputs from our commercial products (e.g. Claude for Work, Anthropic API, Claude Gov, etc.) to train our models”

“we automatically delete inputs and outputs on our backend within 30 days of receipt or generation”

Learn more on the external Website: Is my data used for model training? | Anthropic Privacy Center

No, that would be so-called fine-tuning and we think that this knowledge of which content works well and how it should be designed should lie with the editors, who have a lot of experience and are trained in analysing your web analytics.

Yes, that’s possible, please talk to us, if you wish to have a restricted usage of the Prompts Management UI set up. This could make sense, if you see the Prompts in your news room as a valuable good, as it’s based on a lot of trials and experience, that you want the result to be safe from changes.

If you are interested in having Set up Purple Prompts in your Purple Hub to use it for your Newsroom, just send us a note, we gladly will set it up for you.