TLDR;
Perplexity made its Comet AI browser free. More customers can now use its assistant to research products and act across websites. Brands should check that product comparisons and purchase journeys work, while keeping login and payment safeguards in place.
What happened
Perplexity made its Comet AI browser available free, widening access beyond earlier paid availability. Comet combines browsing with an assistant that can research and act across websites, putting AI interpretation closer to the pages and purchase decisions that customers already encounter.
Why it matters
The browser becomes a place where research and action are delegated. That can expose brittle interfaces, inaccessible details and checkout flows designed around assumptions about human clicks. It also makes indiscriminate blocking a commercial choice with consequences for users who deliberately ask an assistant to help.
Making Comet free lowers the barrier to trying an assistant-led browser, but downloads alone do not establish how buyers will use it. The practical significance is that research and task execution can happen closer together. A customer may ask the browser to interpret a page and then help with the next step, making unclear instructions or brittle interactions a commercial obstacle. For an enterprise brand, successful assistance should mean the intended task is completed correctly with appropriate consent. It should not mean allowing every automated interaction through unchecked. The useful position sits between those two requirements: make public information understandable and ordinary customer actions reliable, while retaining safeguards around identity, payment and account changes. The price announcement is a distribution change; the brand's response should focus on those underlying journey requirements.
How your brand can benefit / be affected
Identify which automated activities the business wants to allow. Test product discovery, comparison and authorised account journeys independently from training crawls. Check whether essential details require unsupported interactions or whether a security challenge blocks legitimate assistance.
Keep authentication and transaction permissions intact. Measure task completion and support failures alongside access logs. A useful agent journey should improve a customer outcome while respecting the same boundaries that apply to other authorised clients.
Pick a few common tasks with a clear end state, such as locating a product's documentation, finding an eligible service option or starting a support request. Review whether the pages explain the required inputs and whether errors help the customer recover. A form that silently clears information or a button with an ambiguous label is a problem for human users too. Where an action changes an account or creates a financial obligation, define the approval the business expects before it proceeds. Do not assume that browser assistance supplies that approval or that the customer understands every action taken on their behalf.
Involve security and customer experience teams in the same review. Look for content that could be confused with operational instructions, unnecessary exposure of account information and journeys where an assistant might continue after the customer intended it to stop. These are assessment questions, not claims that Comet necessarily behaves that way. Log task completion, mistakes and support escalations alongside the time saved. For employees considering the browser, check managed-device requirements and the handling of business information before broad adoption. A useful pilot produces a short list of proven tasks and known limits. A general enthusiasm for agentic browsing is not enough to justify changing access rules or customer processes.
News date: 2 October 2025. Editorial review: 16 September 2026. Analysis includes subsequent developments where stated.