FR: Enforce spelling preferences from .cursorrules at the model level, with hard constraints

Ideas Feature Requests
1

This was previously filed as a GitHub issue before the Issues repository was removed.
Reposting here as requested to preserve the report:

Closed #3805 as not planned.

Cursor does not reliably respect spelling preferences defined in .cursorrules (e.g. Canadian vs American English). Even when explicitly specified, the model frequently reverts to US spellings in generated comments, documentation, and text.

This appears to be a model-level problem, and not something that can be solved consistently through post-processing or user correction (even with many attempts).

Why This Matters

  • Spelling consistency is part of coding and documentation standards.

  • Repeated corrections create review churn and cognitive overhead.

  • Post-hoc fixes don’t prevent errors during regeneration or continuation.

  • Style drift undermines trust in longer AI-assisted writing sessions.

Requested Behaviour

Spelling preferences in .cursorrules should be treated as hard constraints, enforced at the model instruction level.

Concretely:

  • If a language variant (e.g. en-CA) is specified, the model should not emit other variants.

  • Enforcement should be consistent across all generation surfaces.

3

Even when explicitly specified, the model frequently reverts to US spellings

interesting, i just tested this and it actually worked for me. had a .cursorrules with “Always use British English spelling (colour, behaviour, organisation)” and cursor followed it.

might depend on the model or how the rule is phrased? would be curious what exact wording youre using. in my experience the more explicit the better, like listing out specific words you want spelled differently.