Your research starts with a simple question and ends with twelve tabs that seem to disagree. An AI research assistant can help sort those sources into a clear answer, as long as you keep track of where each claim came from. Lensia gives you sidebar chat beside the pages you are reading, with web search and URL fetching on Basic and above. You can ask follow-up questions without moving the whole discussion into another browser tab. Install Lensia free to start with material you supply yourself.
Define the question before collecting more tabs
Research gets easier when the question has a boundary. "Which project management tool is best?" asks for a judgment across almost every team and budget. "Which of these two tools supports the reporting workflow our five-person team uses?" gives you something you can investigate.
State the decision, the constraints, and the kind of evidence you need. For example: "I am comparing two scheduling tools for a small consultancy. We need calendar syncing and a client booking page. Help me identify the information to check on each product's official site."
Ask for a research checklist before a recommendation. That prevents an early suggestion from steering every later search. Include questions that might disqualify an option, not just features you hope to find.
If you are reading to understand an unfamiliar topic, define the scope differently: what you already know, which terms are confusing, and what you want to explain by the end. You might simply want to understand what you are reading.
Separate supplied material from web research
Lensia's sidebar supports conversations about material you paste or attach. Authenticated chat can also read available text from the current page. Web search and fetching a supplied URL are available on Basic and above. Those tools are distinct from the Pro plan's opt-in Agentic Mode.
You can start with Free by supplying a short passage, then asking it to explain the author's claim. If you need live sources, compare the paid plans and choose one with search access.
Tell the assistant which mode you want: "Use only these excerpts" or "Look for current official documentation, and identify any information you add from outside my notes." Source boundaries matter because the assistant can otherwise combine your material with general knowledge in a way that is hard to untangle.
An answer about a current price or release needs current evidence. Ask for the date of the source, then visit the page to verify it. A search result snippet can be older than the page it points to.
Build a source list you can inspect
- Write your research question and constraints in sidebar chat.
- Supply the pages or excerpts you already have.
- On a paid plan, ask for relevant current sources where needed.
- Open each important source and confirm it supports the claim.
- Record the title, link, date, and useful passage in your notes.
- Ask the assistant which parts of your question remain unanswered.
Prefer the organization responsible for the information when that organization can answer the question directly. Product documentation is a good place to check a feature; a vendor's marketing page is less useful for deciding whether customers find it easy to use.
Use independent sources when the question requires them, and keep their role clear. A user report may reveal a problem worth investigating, but one person's experience does not establish how often it happens. Evidence quality depends on the claim you are trying to support.
Ask for comparisons that retain disagreements
Suppose two sources give different explanations for the same change. Asking for a single smooth summary can hide the conflict. Try: "Compare these sources. Identify what they agree on, where they disagree, and whether they cover the same dates or population. Do not resolve a disagreement without evidence."
A conflict may turn out to be a difference in scope. One source discusses a trial; another describes the general release. One includes an optional add-on; another describes the base plan. Put those differences into the comparison before deciding that either source is wrong.
For documents, label the filenames and versions. Our PDF comparison guide gives a focused workflow for keeping those details straight. For a long online source, use the article summary prompts to extract the evidence before comparing it.
Turn the conversation into a brief
Write for a specific reader
Ask for a brief that names the question, the supported findings, and the next step. Tell the assistant whether the reader already understands the topic. A colleague choosing between two options needs different context from someone learning the vocabulary for the first time.
Try: "Draft a short brief for a teammate. Include our question, findings supported by the sources, unresolved issues, and the next check we should make. Keep source links beside the claims."
Preserve uncertainty
Unanswered questions belong in the brief. If a source does not specify a limit, mark that gap instead of treating it as unlimited. If your information is only current as of a particular date, retain that date.
Review every link before sharing the brief. The assistant may attach a real source to a claim that the source does not support. If you cannot find the supporting passage, revise or remove the claim.
Our email writing guide can help you turn the checked brief into a message without adding promises or facts.
Use models and browser actions deliberately
Lensia includes a model picker, so you can try a different model when an explanation stays unclear. Keep the same question and sources when comparing replies. Changing the evidence and the model together makes it difficult to tell what improved the answer.
Agreement between models is not independent confirmation. Both can make the same mistake. Use the source itself to settle a factual question; use another model to explore an explanation or identify a possible gap.
Pro's Agentic Mode can carry out browser tasks with multiple steps when enabled. It is optional, and it follows its own permissions. For a simple reading comparison, ordinary sidebar chat may be enough. Start with a manageable scope rather than asking an agent to "research everything" about a topic.
Finish with the next decision
Before you collect another source, ask what information would change your conclusion. If none of your open questions affects the decision, you may already have enough to act. If one does, make it your next search.
A useful research session ends with an answer you can explain, the evidence behind it, and a short list of remaining checks. You can bookmark a helpful sidebar response and keep the final brief with your project notes.
Start researching with Lensia: bring one question, two sources, and a reason you need the answer.