How to Use Perplexity Pro Search and Choose the Right Model

Pro Search is the Perplexity mode aimed at harder multi step questions that need more thorough retrieval than everyday Search. You still get live web grounded answers with numbered citations, but the path through sources can be deeper. Model selection on eligible Pro access lets you steer style and reasoning. For the product map, begin at /explore/perplexity.

This guide covers when to choose Pro Search, how Free and Pro plans differ at a high level, how to think about Best mode and named models, and a practical workflow. Pair it with Research reports at /blog/how-to-use-perplexity-research-for-deep-reports and with fast Search habits at /blog/how-to-use-perplexity-for-fast-web-search-and-better-answers.

When to use Pro Search

Choose Pro Search when the question chains constraints, when you need several retrieval steps, or when a shallow answer would miss tradeoffs. Examples include comparing plans across vendors with region rules, rebuilding a timeline from conflicting news, or extracting requirements from mixed public docs. Choose Search for quick facts. Choose Research for long multi source reports with a clear brief.

Free users get Search with citations and limited Research and Pro Search. Pro is listed at $20 per month on https://www.perplexity.ai/hub/pricing and extends Research and Pro Search while adding model selection. Max is listed at $200 per month at a high level. Do not invent daily quotas. Confirm live limits in Perplexity help.

How Pro Search and models fit together

You pick a mode from the selector on web, mobile, and the Mac app. Pro Search runs a more thorough multi step pass when your access allows it. Threads keep context for follow ups. Help also mentions Best and Reasoning Search. Treat those as product labels that can change, and trust the live selector over memory.

On Pro, Best mode can auto pick a model. You can also choose from models shown in your account. Examples that have appeared in product help include GPT 5.2, Claude Sonnet 4.6, and Gemini 3.1 Pro among others. Model menus change. Always check the live list before you standardize a team process on one name.

Developer API access at https://docs.perplexity.ai is billed with separate credits. Mention it only when your team needs programmatic search. This article focuses on the consumer Pro Search and model picker experience.

Step by step Pro Search workflow

1. Write the multi step job clearly

List the steps you need completed. Pro Search works better when the chain is explicit instead of implied.

Job steps:
1) Identify official pricing pages for [Vendor A] and [Vendor B]
2) Extract publicly listed mid tier plan prices for [region]
3) Note context window or seat minimums if stated
4) Flag any contact sales only plans
Output: comparison table with citation numbers

2. Add constraints and source priorities

State exclusions early. Tell Perplexity to prefer primary pages. Ask it to mark conflicts instead of averaging them away.

Constraints:
Region: United States public pricing only
Prefer: official pricing and docs pages
Exclude: affiliate roundups and unverified screenshots
If sources conflict: show both claims with citations
Do not invent numeric limits

3. Choose mode and model intentionally

Select Pro Search for the hard chain. Use Best if you want automatic model choice, or pick a listed model when you already know which style you need. Re check the live model list when onboarding new teammates.

Mode: Pro Search
Model: Best (auto pick) unless my team standard says otherwise
Prompt pattern: instruction, context, input, keywords, output format
Require numbered citations on every factual cell in the table

4. Run, then verify the hard claims

Open citations for prices, limits, and dates. For hallucination risk patterns see /blog/reducing-hallucinations-with-perplexity. Summaries can sound complete while a citation points to a marketing page that never stated the number.

Verification checklist
[ ] Every price opened on a primary page
[ ] Region matches the constraint
[ ] Contact sales plans labeled correctly
[ ] Conflicts listed instead of hidden
[ ] No confidential internal data was pasted

5. Follow up for gaps only

Stay in the thread. Ask for missing vendors, a tighter table, or a one page brief. Avoid regenerating the entire answer unless structure failed.

Follow up:
Add [Vendor C] using the same columns.
Remove any row that lacks an official pricing citation.
Produce a 5 bullet exec summary above the table.

Prompt library for Pro Search jobs

These prompts assume Pro Search or another deep mode on eligible access. Keep Free plan limits in mind when you batch work.

Cross vendor constraint solver

Find tools that meet all constraints:
Must have: SSO SAML, EU data residency option stated publicly, export to CSV
Nice to have: audit log export
Return only tools with official pages for each must have.
Table: Tool | Must haves met | Gaps | Citations

Conflicting news timeline

Build a timeline for [event] across the last 30 days.
Separate confirmed, contested, and rumor.
For contested items list competing outlets with citations.
End with what a careful reader still cannot know.

Requirements extraction from mixed docs

From official docs for [product], extract requirements to enable [feature].
Steps: find the feature page, list prerequisites, list admin permissions, list known limits.
Cite each prerequisite. Mark anything undocumented.

Model choice experiment note

Run the same prompt twice conceptually for my notes:
Prompt: [paste]
Model A: Best
Model B: [named model from live list]
I will compare citation quality and structure myself.
Remind me to verify both against primary pages.

Procurement packet

Build a procurement packet summary for [category].
Include: shortlist of 5 vendors, public pricing signals, security page themes, support channels.
Cite official pages. List questions for sales that public pages do not answer.

Regulation impact matrix

For [regulation] impacting [industry] in [region]:
Matrix columns: Obligation, Who it applies to, Effective timing if stated, Primary source citation
Prefer government texts. Label secondary explainers clearly.

Technical migration risk scan

Scan public docs for migrating from [system A] to [system B].
Cover: official migration guides, breaking changes, quota or rate limit pages, rollback notes if any.
Cite docs. Do not invent downtime estimates.

Executive decision brief

After the Pro Search answer, produce an exec brief:
Decision asked
3 options with citation backed tradeoffs
Recommended option with one paragraph why
Open diligence items for next week

Source tier rewrite

Re rank sources used in the prior answer.
Tier 1 primary official pages
Tier 2 reputable news or peer reviewed
Tier 3 blogs
Rewrite claims that rest only on Tier 3 so they are removed or clearly labeled.

Weak vs strong Pro Search ask

Weak: Research the best AI coding tools.
Strong: Pro Search compare public pricing and stated context limits for major coding assistants updated in the last 6 months.
Table: Product | Free tier signal | Paid entry price | Notable limits | Citations
Official pricing pages only.

Verification and quality control

Pro Search can feel authoritative because it is thorough. Thorough is not the same as verified. Open primary pages for any claim that affects spend, compliance, or customer promises. Structured formatting help lives at /blog/how-to-get-structured-and-formatted-answers-from-perplexity.

• Match mode to job: Search for speed, Research for long reports, Pro Search for hard chains
• Re check the live model list before documenting a team standard
• Confirm Free versus Pro expectations on the official pricing page
• Keep Max mentions high level at $200 per month until you read the live inclusions
• Preserve citation links in the notes you share with teammates

Document which jobs your team sends to Pro Search versus Search or Research. Clear routing reduces wasted deeper runs and keeps Free plan limits from surprising people mid project.

When you evaluate models, hold the prompt fixed and compare citation usefulness and structure, not just prose style. Style differences are easy to notice. Source quality differences matter more for decisions.

Common mistakes

• Using Pro Search for a one fact lookup that Search would finish faster
• Assuming Free access matches Pro depth without checking help
• Standardizing on a model name without confirming it still appears in product
• Skipping citation opens because the multi step answer looked complete
• Hiding source conflicts instead of listing them
• Pasting confidential roadmaps into a web grounded prompt
• Regenerating whole answers instead of targeted follow ups

Related articles: /blog/how-to-use-perplexity-research-for-deep-reports, /blog/how-to-use-perplexity-for-fast-web-search-and-better-answers, and /blog/reducing-hallucinations-with-perplexity. Overview: /explore/perplexity.