Overview
A research-first AI search engine that keeps sources visible while answering questions.
Key features
- Every answer includes numbered source citations
- Pro Search runs multi-step research instead of one-shot answers
- Focus modes: academic, Reddit, YouTube, news
- Spaces for organizing research threads by project
- Model switching inside a single product
Trial checklist
Use a real task involving competitor scans—not a demo prompt. Verify whether every answer includes numbered source citations materially reduces editing or setup time. Record output quality, time to a publishable result, and total plan usage. Pay particular attention to sources still need checking so factor that into your monthly budget. Repeat the same task in at least one alternative before deciding.
Use category-specific evidence
Arena Text Leaderboard [1] ↗ · LiveBench [4] ↗ · Artificial Analysis methodology [5] ↗
Arena Text is a human-preference leaderboard from anonymous pairwise battles. Its rank is the model's position; its rating is the statistical preference score; and any ± value is a confidence interval, not a guarantee of quality. See benchmark policy →
Pricing
Current official summary: Official support lists Standard/Free, individual Pro, individual Max, Enterprise Pro, and Enterprise Max. Max includes everything in Pro plus the highest access to advanced models, Comet Max Assistant, Create files and apps, and Brain. The reviewed public consumer-plan pages did not expose current Pro or Max dollar prices. An official organization-onboarding article lists Enterprise Pro at $40/seat monthly or $400/seat annually.
Open verification questions
- Current public prices for individual Pro and Max were not verifiable from the reviewed official pages.
- A blanket commercial-use or output-ownership claim was not made because consumer outputs may be governed by third-party model licenses.
- Enterprise Max pricing was not publicly stated in the reviewed sources.
- The official Lily blog says the chat-completions API streams tokens; the official lily README in pplx-garden as of 2026-09-02 documents stream: false and rejects streaming. Treat serving-surface details as demo-specific until re-checked.
- Lily hardware support is stated as Apple GPU family 10 or later (M5 and newer) and macOS 26 or later in the official demo README; support on every later Apple chip or OS was not independently verified.
Perplexity is documented as an AI-powered answer/search product with citations, premium models, file uploads, Research, Create files and apps, Comet, and Computer features depending on plan. AI Data Retention is enabled by default for Free, Pro, and Max users and can be turned off for future collection; previously collected training data is not removed by a later opt-out. Enterprise data is not used for AI training, and Perplexity says its third-party model-provider agreements prohibit provider training on Perplexity data. Consumer terms warn that output may be subject to third-party model licenses and use restrictions. 2026-09-02 official blog “Optimizing On-Device Inference for Apple Silicon” (https://www.perplexity.ai/hub/blog/optimizing-on-device-inference-for-apple-silicon): Lily is a lightweight local inference engine built specifically for Apple silicon and Qwen3.6-35B-A3B so on-device compute does not bottleneck Hybrid Compute / Perplexity Computer tasks. Official facts from that post: Rust runtime; OpenAI-compatible chat-completions API; custom Metal kernels; neither PyTorch nor MLX in the execution path; standalone demo on GitHub. The official pplx-garden README names the demo at https://github.com/perplexityai/pplx-garden (lily/), Apache-2.0, greedy generation for the Qwen3.6-35B-A3B MLX affine 4-bit checkpoint format, Apple GPU family 10 or later (M5 and newer). Lily is not a new paid Perplexity plan SKU. Consumer Pro/Max dollar prices were not re-verified on 2026-09-02. Re-verified 2026-10-01: the official plan-comparison article still routes consumer Pro and Max to the app rather than publishing dollar amounts, and lists Education Pro at $10/mo with SheerID verification. Enterprise Pro is stated as starting at $40/month or $400/year per seat, matching the 2026-08-25 record.
Pricing changes often — verify current plans on the official site before buying. Dated snapshots: Pricing Watch.
Use cases
- Competitor scans
- Source discovery
- Market questions
- Briefing prep
- Vetting sources with citations for reports
- Tracking competitor moves across the web
Pros
- Citations are central
- Good for discovery
- Fast current research
Cons
- Sources still need checking
- Less suited to complex workflows alone
Who Perplexity is for
Perplexity is aimed at teams and individuals who need source-backed web research and quick market scans. It is designed for research workflows where output quality, adoption effort, and operating cost all matter. Compare alternatives first if you require self-hosting, unusually deep customization, or tighter policy controls than the official plan provides.
Buying decision
| Keep it when | Source-backed web research and quick market scans is the job you repeat. |
|---|---|
| Test first when | Sources still need checking. |
| Before you pay | Run the trial checklist and check the current plan limit on the official pricing page. |
Verdict
The call: Perplexity has an AIToolsEssentials editorial score of 4.6/5. Citations are central. Price this before you renew: sources still need checking.
Run one real task from your work this week. That tells you more than this review.
How Perplexity compares
In research, Perplexity sits next to NotebookLM, You.com, CrowdWisdomTrading. The useful part is citations are central. The part to price first is sources still need checking. Read the side-by-side before you pay for a second seat.
How we evaluated
The AIToolsEssentials rating is an editorial score—not an external benchmark. It summarizes job fit, likely output quality, ease of adoption, and operational cost using published product information, benchmark context where the exact model is identifiable, and the repeatable trial checklist above. Benchmarks never determine the final product rating by themselves. See our editorial methodology, evidence ledger, and benchmark evidence policy.
Frequently asked questions
Is Perplexity better than Google?
For research questions where you need sources, yes — it synthesizes instead of listing links. For local searches and shopping, Google still wins.
Does Perplexity use ChatGPT?
It offers multiple models including GPT-4o and Claude alongside its own Sonar models — you choose per search.
Can I trust the citations?
Mostly. Every claim links to a source so you can verify, but we found it occasionally picks low-authority pages on obscure topics.
Cite or embed this audit
The badge links to this review. It is a dated price receipt, not a ranking and not a paid placement.
