Perplexity deprecated direct DALL-E 3 selection and the new workflow is worth understanding before you need it

Direct selection of DALL-E 3 within Perplexity on paid Pro plans has been removed. The platform now uses automated internal routing that chooses the image generation model based on your prompt without disclosing which model it used.

The trigger phrases, generate an image, create an image or make a picture, are what activate the image generation pathway. Without those phrases the response stays as text. You are not controlling the model. You are controlling the prompt quality.

Keeping your model selection on default text options like GPT-5.4 or Gemini 3.1 Pro and relying on detailed prompts with explicit style, mood, colour and quality descriptions is the new workflow.

The core priority being search and research rather than image generation is the Perplexity product decision that makes this change legible. It is not a degradation of a core feature, it is a deprioritisation of a peripheral one.

For anyone using Perplexity primarily as a research tool who also needs occasional image generation: has the quality from automated routing been acceptable or are you now switching to a dedicated image tool for that part of your workflow?