Conversational answers are no longer the whole proposition. Search products combine research, personalisation, sources and tools in different ways. This analysis describes selected official materials checked on 29 October 2026, not an exhaustive feature audit or independent performance benchmark.
Six views of the market.
Availability varies by plan, region, account and rollout. A feature appearing in official material does not establish that every user has it. Absence from this summary does not mean a competitor lacks a capability.
Perplexity
Memory can carry preferences across conversations, with controls and private-session options. Projects organise ongoing work with files, instructions and tools. Deep Research and multi-model features broaden the research workflow.
Implication: persistent context and research workspaces are already competitive baseline features.
Deep research brings a reviewable research plan, sources and reports together. Agent capabilities extend the product into tool-assisted tasks, connected sources and scheduled work.
Implication: the path from a research question to useful work is already being built by major platforms.
Google’s Personal Intelligence announcement describes opt-in use of connected Gmail and Photos context within AI Mode. Its relevance is the combination of search distribution and personal context; current access conditions must be checked separately.
Implication: “search that knows you” needs a clearer proposition than personalisation alone.
Custom Agents support model choice, instructions, source controls, personalisation and research settings. Tool documentation also describes search and research capabilities for agent workflows.
Implication: configurable specialists and tools are established product patterns.
Ask Brave combines search and conversational follow-ups with web, video and product results. Brave also operates an independent search index and offers search data through an API.
Implication: trustworthy retrieval and a clear privacy approach matter alongside the conversational interface.
Copilot documentation describes memory, chat-history context and work insights. Microsoft’s support notice says consumer Deep Research began retiring on 18 August 2026, while Microsoft 365 Premium subscribers can use the separate Researcher experience.
Implication: product names and packaging change; comparisons need dates and specific product boundaries.
The market suggests a common set of expectations: natural-language questions, follow-ups, cited answers, deeper research modes, reusable context and tools that connect research to work. These expectations are an inference from the selected sources, not a statement that every product offers every feature.
For ask.pa, the useful question is how those capabilities support a complete research thread: what was asked, which context was used, what evidence supports the answer, what remains uncertain and what the user approved next.
Ten priorities, not ten exclusivity claims.
The ten directions on the homepage form a proposed product strategy. Related functions may exist elsewhere. The opportunity is a coherent, inspectable workflow, and its value needs user testing.
Evidence expiry: do review dates prevent stale decisions?
Contradiction ledger: can users resolve disagreements more quickly?
Decision replay: can teams see why a recommendation changed?
Knowledge debt: do explicit gaps reduce repeated research?
Context firewalls: can people reliably keep client contexts separate?
Research budgets: can users control effort without losing important evidence?
Source independence: can the system distinguish corroboration from repetition?
Unknowns queue: does the next question improve the decision?
Reversible handoffs: can users approve a proposed action with enough context?
Research passport: can another person or tool resume the work accurately?
Only a small curated research workbench and browser-memory demonstration are implemented on this site. The wider priorities are product directions, not promised production capabilities.
Earn the advantage.
Start with a group that repeats consequential research: procurement teams, founders or client-facing advisers. Compare a bounded ask.pa workflow with their existing process using the same questions and evidence requirements.
Measure completion quality and time, including the reviewer’s effort.
Check citation support, freshness and source independence manually.
Test incorrect context, conflicting sources and missing information.
Ask users to explain what was remembered and how to remove it.
Measure whether a second researcher can resume from an exported record.
No customer study, revenue claim, benchmark result or proprietary search index is asserted here. A compelling direction becomes a business through validated usefulness, reliable execution and clear economics.