Kagi Search Lenses: Scope Sources, Then Check the Evidence

Suppose a founder wants to know whether a shared research configuration stays synchronized after someone copies it. Official documentation should settle the product behavior. Whether that configuration suits the team's research is a different question. A search restricted to official sources can help with the first and leave the second unanswered.

That distinction matters more than the number of search controls you switch on. Kagi search lenses save source boundaries; personalized results set ongoing website preferences. Filters and operators handle individual queries. Use the control that fits the question, then read the evidence it helps you find.

The setup below is based on public documentation. Its queries are hypothetical examples, not results from a logged-in Kagi session or a search-quality benchmark.

Choose a lens, a domain preference, or a query filter

A reusable source boundary is not a ranking preference

A Kagi lens saves parameters that constrain searches, including selected websites. Personalized results change how websites are treated in future searches. The options are Block, Lower, Normal, Higher, and Pin. They express website preferences, not weights assigned within a lens.

For a single search, filters or documented search operators may be enough. site:help.kagi.com restricts results to that website. filetype: selects a file extension; inurl: and intitle: look for terms in URLs and titles. Quoted phrases search for the exact words. Adding restrictions is worthwhile only if you can explain what each one contributes.

A compact choice table for recurring and one-off research

To suspend personalization for one search, uncheck Personalized in Options. Record whether you did. Turning off a lens does not mean you also turned off personalization. Otherwise, a supposedly broader search may still carry a website preference you meant to set aside.

Build a small docs-first lens for a recurring question

Define the question and the source boundary

Take the founder's question: after someone imports a shared Kagi lens, do later changes reach their copy? For this hypothetical setup, name the lens Kagi official docs and include help.kagi.com. That is a sensible boundary for finding documented behavior. It cannot tell you whether the resulting workflow meets your team's needs.

Leave additional constraints off initially. A date cutoff, for example, could hide useful context without helping establish how sharing works now. Add a restriction only after identifying a reason for it.

The lenses documentation permits up to 10 included and 10 excluded websites, plus up to 5 included and 5 excluded keywords. It also offers region, file type, and before/after date settings. Lenses do not support image or video searches. These limits describe the controls available; they say nothing about the quality of the evidence returned.

Save, enable, and inspect the lens

Follow the documented sequence in Lenses settings. Choose Create New, enter the name and included website, then Save. Switch the saved lens to On in the list. Run a search and enable the lens on the results. This is a procedure for the reader to try, not an account change made for this article.

Before opening results, inspect the active controls and visible domains. Sparse results are a prompt to revisit the boundary or query wording. They do not prove there is no answer.

Preset names deserve the same scrutiny. The Programming preset includes official programming-language websites and forums. Kagi's Documentation preset covers help, blog, and feedback sites. Neither has the same scope as this narrower official-docs example. Read the configuration before relying on the label.

Use the Scope, Contrast, Verify research playbook

Scope: answer the documented behavior question

The Optijara Scope, Contrast, Verify playbook is an editorial framework proposed in this guide. It is not a Kagi feature or a validated method. The point is to stop a source-selection decision from quietly becoming a judgment about truth.

With the hypothetical docs-first lens enabled, try Kagi lenses sharing updates. Open the official page and find the sharing section. Opening a sharing URL loads the original lens's latest settings at that moment. Saving the imported configuration creates a copy that does not receive later updates. Kagi documents both behaviors. A useful answer must preserve the distinction.

Contrast: search for practical limits outside the lens

Remove the lens and record whether personalization remains active. A broader query such as Kagi lenses included sites weighting could help uncover questions about practical use. It is a discovery prompt, not evidence of a defect. Quick Answer can trigger automatically for queries ending in ? when Auto Quick Answer is enabled. Leave off that trailing mark, or disable Auto Quick Answer, if you want to inspect results first.

A forum request needs context. Check its date and distinguish a requested behavior from a currently documented feature. An old complaint can suggest something to investigate without establishing what the product does now. Equally, an official page can explain a control without settling whether you should adopt it.

Be specific about the gap you are trying to fill. If the question concerns documented behavior, return to primary documentation. If it concerns suitability, write down your requirements first. A practitioner account may identify a condition worth testing, but it is not automatically representative. Where accounts conflict, keep both passages and their contexts. Do not resolve the disagreement by choosing the one you prefer.

Verify: connect each conclusion to a source passage

Open the candidate page and identify its publisher. Read the relevant passage with its qualifications, rather than stopping at the search snippet. Note any stated update date or applicable version. If neither appears, say so. Your interpretation should be distinguishable from the source's wording.

A useful handoff check is to ask another reader to reopen the passage without your explanation. Can they find the qualification that matters? Can they tell which part is documented behavior and which part is your recommendation? If not, improve the note. Record your own consultation date separately from the publisher's date; neither promises that the page will remain unchanged.

Use the following checklist before accepting a finding.

Keep the canonical URL alongside the question and relevant passage. The note also needs your interpretation, any stated date or version, and whatever remains unresolved. For a longer writing project, the guide to keeping research notes connected to their sources develops this source-to-sentence handoff.

PDF tables need extra care. Retain headers and units with the value, plus the document revision and any conditions attached to it. The guide to preserving source references when extracting PDF evidence covers that separate extraction problem. It is a step after discovery, not another Kagi control.

If you repeat this workflow, keep a small review log. Record where conclusions changed and why, rather than announcing an accuracy improvement you have not measured.

The JSON below summarizes the proposed framework. It is not a Kagi API payload or a completed search record. No execution result is implied.

{
  "framework": "scope-contrast-verify",
  "stages": ["scope", "contrast", "verify"],
  "source_boundary": "official-docs-then-broader-discovery",
  "evidence_note_fields": ["question", "source_url", "passage", "date_or_version", "interpretation", "unresolved_limit"],
  "review_rule": "revise-or-omit-unsupported-sentence",
  "execution_status": "illustrative-unexecuted"
}

Use Quick Answer as synthesis, not source approval

What the documented references establish

Quick Answer summarizes results across returned pages and provides references to the pages used. Once results appear, select Quick Answer or press q. If Auto Quick Answer is enabled, a query ending in ? triggers it automatically. Auto Quick Answer can be disabled in Search AI settings.

The reviewed documentation does not establish that every synthesis respects every lens and personalization restriction. References do not supply that missing guarantee. For a product-capability question that will inform a decision, read the source pages before asking a summary to organize the answer.

Read the source before accepting the sentence

Consider the hypothetical summary “shared lenses stay updated.” It sounds clear, but it loses the distinction between opening the original sharing link and using an imported copy. That sentence needs rewriting even if a reference points to the right documentation. This is a review example, not a Quick Answer response observed during authenticated search.

A reference tells you a page was used. It does not tell you whether the relevant passage supports every part of the sentence. Look for qualifications the summary may have dropped, and check whether the passage addresses your actual question. Keep recommendations separate from statements about product behavior.

If the support is missing, narrow or remove the sentence. Leave a conflict unresolved when the evidence cannot settle it. An honest gap is easier to work with than a confident answer that requires the reader to overlook its source.

Share lenses with copy and update expectations intact

A share link loads settings; an imported lens is a copy

The documented sharing procedure starts by editing the lens, enabling Share Lens Link, and saving. Use the sharing icon to copy the URL. A recipient with a Kagi account opens a prefilled Create Lens form and can edit it before saving.

Opening the link loads the original lens's current settings. An existing copy does not receive subsequent updates. Disabling sharing invalidates the old link; re-enabling it generates a new one. Calling this centrally synchronized configuration would give recipients the wrong expectation.

Attach a short scope and review note

Include a brief handoff note with the lens. Explain its purpose and source boundary, including what you expect it to omit. Choose a review date and say how recipients should compare or re-import settings when a change matters. These are suggested communication practices, not built-in notifications or enforcement.

Invalidating a sharing link does not establish that recipient copies or previously recorded information have been deleted. Keep private project details out of the example configuration, and limit revocation claims to the documented link behavior.

Common mistakes and the limits of source selection

Narrowing discovery until nothing disagrees

A lens is a poor substitute for judgment. The mistake is treating a familiar set of domains as permission to stop reading critically. If a conclusion outruns its source, adding another preferred domain does not fix it. Rewrite the conclusion.

Official documentation alone also cannot settle a broader adoption decision. It can establish a control's described behavior. Suitability depends on your requirements, with a separate search for practical limits. A broader pass helps look for omissions; it does not prove balanced or exhaustive coverage.

Other mistakes are more mechanical. Assuming Programming means official-only changes your understanding of the boundary. Treating a shared copy as live configuration changes your expectations about maintenance. Inspect the preset, and tell recipients how updates will be handled.

Confusing filters, domain controls, and evidence quality

Recency ordering and date filters change retrieval, not factual currency or release applicability. Inspect the page's dates and context. Kagi's built-in domain blocking does not support path wildcards or regular expressions. A domain preference should not be mistaken for a path-sensitive source policy.

This guide draws on public documentation. It reports no paid-query trials or measured search-quality results, and offers no client deployment evidence, pricing comparison, or time-saving benchmark. Reviewing sources and maintaining boundaries take effort. Whether that effort is worthwhile depends on the question and the available evidence; no ROI or performance improvement has been established here.

Keep confidential project details out of example queries. This guide does not audit Kagi's data retention or certify privacy. Start with a non-sensitive behavior question. The result worth keeping is a conclusion another reader can trace back to the passage that supports it.