There are so many layers to this question!
First, we need to figure out what "cognitive effort the assessment is intended to elicit", as you write. That is something that is extremely context-dependent and that you need to figure out together with your colleagues.
Then, we need to look at (and ideally talk with our students about) how GenAI is actually being used. Students in your course might be reluctant to have a honest conversation about their GenAI use with you, but maybe you can also ask students that took your course already and are not in a power relationship with you any more, or ask colleagues elsewhere what they know about how their students use GenAI? There is also a lot of survey data available online which might be useful as a first indication of what to expect, but you might want to try to find more context-specific information.
Then, we need to look at the different ways GenAI is being used. Are there some that you think are removing the cognitive effort you want? Are there others that you think support learning? Do your students agree with your assessment? Are there some uses that you are unsure about and that you need to put on a "keep an eye on and revisit" list?
And then, lastly, we can look at the uses and our evaluation of which support learning and which take away effort that you want to see in assessment (and that might be different from effort that you want to see at some stage of the learning process, for example spelling might be a learning goal at some point, whereas later that is 100% outsourced to spellcheckers), and write that into a GenAI guidance. Personal recommendation for a great tool to support the writing process: The GenAI Guidance Wizard by Rachel Forsyth on CanAI.education!
P.S.: Remember that GenAI use is a very slippery slope. Even students who know that it harms their learning can slide into GenAI dependence, for example as an anxiety response. So the GenAI guidance you provide is important, but it is at least equally as important that students are given the opportunity and support to practice emotional regulation and to learn to protect self-efficacy under stress...
---
Wu, H., Ni, H., He, J., & Wee, E. H. (2026). AI anxiety and AI dependence among undergraduates: a moderated mediation model of AI self-efficacy and AI literacy. Frontiers in Psychology, 17, 1884382.