The most reliable figure for Perplexity AI monthly queries remains 780 million searches in May 2025. Chief executive Aravind Srinivas disclosed that number at the Bloomberg Tech conference in San Francisco on June 5, 2025, adding that query volume was growing by more than 20 percent month over month. He also set an ambitious goal of reaching one billion queries per week by the end of 2025.
As of June 11, 2026, however, Perplexity has not released a comparably specific, companywide monthly-query figure confirming that the target was reached. Claims that the platform now processes 2.1 billion, four billion or more queries per month are estimates rather than audited disclosures. Figures claiming 60 million to 70 million daily searches, 100 million monthly users or 480 million monthly Comet searches cannot be treated as official without primary documentation.
There is also an important mathematical correction. A platform handling 780 million queries per month is running at an annualized pace of approximately 9.36 billion queries, not 52 billion. The figure of roughly 52 billion per year corresponds to Srinivas’s separate target of one billion queries per week.
That distinction matters because Perplexity in 2026 is no longer merely a question-and-answer website. Its workload now includes Pro Search, Deep Research, API searches, Comet browser actions, connectors and multi-step Computer assignments. A single user instruction can trigger many internal searches, model calls, URL fetches and tool operations. Consequently, the meaning of a query is becoming less stable precisely as the company’s infrastructure becomes more important.
Perplexity AI Monthly Queries: What Is Officially Verified?
Perplexity’s verified query-volume history is narrower than many statistics pages suggest. The strongest public benchmark is the May 2025 disclosure of 780 million queries, representing more than 20 percent growth from April. Contemporary reporting also placed the company’s ambition at one billion weekly queries by the end of 2025.
The May figure converts into approximately 25.2 million queries per day over a 31-day month, 176 million queries per week, 300 queries per second when averaged continuously, and 9.36 billion queries on a simple annualized basis. Those calculations describe average traffic, not peak capacity. Search traffic is uneven by hour, geography and product.
The reported 20 percent monthly growth rate should not be projected indefinitely. Compounding 780 million by 20 percent for 12 months would produce nearly seven billion monthly queries by May 2026. No official Perplexity disclosure currently confirms such a figure.
Table 1: Perplexity AI Verified Query Metrics (May 2025)
| Query Metric | Value | Status |
| Queries in May 2025 | 780 million | Official executive disclosure |
| Month-over-month growth (May 2025) | More than 20% | Official executive disclosure |
| Average daily volume (May 2025) | ~25.2 million | Calculated |
| Average weekly volume (May 2025) | ~176 million | Calculated |
| Annualized May 2025 run rate | 9.36 billion | Calculated |
| Target announced for end of 2025 | 1 billion per week | Executive target, unconfirmed |
| Annual pace at target level | ~52 billion | Calculated from weekly target |
| Official 2026 monthly total | Not publicly disclosed | Unconfirmed |
The Critical Difference Between a Target and a Reported Result
The one-billion-query weekly target is frequently presented online as though Perplexity achieved it. The original statement was forward-looking. Srinivas said the company was aiming for that level by the end of 2025, based partly on the growth rate visible in mid-2025. One billion searches per week would equal roughly 142.9 million per day and approximately 4.35 billion monthly queries — 5.6 times the May 2025 level.
Table 2: Mechanical 20% Monthly Growth Projection from May 2025 (Scenario Only — Not a Forecast)
| Month | Projected Queries at 20% Monthly Growth |
| June 2025 | 936 million |
| July 2025 | 1.12 billion |
| August 2025 | 1.35 billion |
| September 2025 | 1.62 billion |
| October 2025 | 1.94 billion |
| November 2025 | 2.33 billion |
| December 2025 | 2.80 billion |
| January 2026 | 3.35 billion |
| March 2026 | 4.83 billion |
| May 2026 | 6.95 billion |
This table is a mathematical scenario, not a forecast or reported operating result. Growth rates normally decelerate as the comparison base expands. Capacity constraints, acquisition costs, product limits, seasonality and user retention all prevent early-stage percentage growth from continuing unchanged.
Why the 2026 Perplexity AI Monthly Queries Figure Is Uncertain
Perplexity has expanded the definition of usage without publishing a standardized query-count methodology. A conventional user question might create one visible query. A Deep Research assignment may search hundreds of sources. A Computer workflow may run for hours, use multiple models, retrieve numerous URLs and repeatedly call connected applications.
Perplexity’s current developer platform contains four major API families: the Agent API for third-party models and tools; the Search API for ranked web results; the Sonar API for citation-grounded generated answers; and the Embeddings API for semantic vectors. Each product can count activity differently. An API request is not necessarily equivalent to a consumer search. Unless Perplexity defines whether its metric counts user prompts, retrieval calls, generated answers or all search operations, comparisons across years will remain imperfect.
From Answer Engine to Agentic Workload
Perplexity’s 2026 product strategy changes the economics behind query volume. In February, the company introduced Perplexity Computer, describing it as a system that combines research, analysis, design, coding, deployment and task management. Srinivas wrote that Computer can orchestrate more than 20 models and decide which model should handle each portion of a job.
“Computer unifies every AI capability in the market into one.” — Aravind Srinivas, Perplexity CEO, March 2026
The statement signals a shift from short interactive searches toward long-running work units. A request such as “compare three vendors, review their contracts and create a recommendation” can produce multiple web searches, page fetches, document analyses, calculations and model responses. Perplexity’s March 2026 product update also added custom remote connectors through the Model Context Protocol, more than 400 curated connectors, and a Snowflake integration that constructs a semantic Data Map from schemas, tables, column relationships and query history.
Perplexity’s 2026 Product and Feature Stack
Core Consumer Features
According to the latest 2026 documentation reviewed, the consumer interface includes standard searches, multi-step Pro queries, Deep Research reports, file analysis, asset generation, video generation, private Spaces, model selection, memory and access through desktop and mobile applications. The official pricing comparison lists up to 200 Pro queries per week and up to 20 Deep Research queries per month for an individual Pro account, along with up to 25 generated assets and three videos per month. Files submitted for answer generation must be smaller than 50 MB.
Enterprise Features
Enterprise Pro adds organizational controls and internal search across web sources, files and workplace applications. Documented capabilities include guaranteed exclusion of customer data from model training, single sign-on, SCIM provisioning, user permissions, premium sources such as PitchBook and Statista, higher file allowances, dedicated support, shared Spaces, workplace connectors, and SOC 2 Type II, HIPAA, GDPR and PCI DSS compliance. Enterprise Max adds larger datasets, greater upload limits, high-end reasoning models, Model Council comparison, broader Deep Research capacity, audit logs, configurable retention and team analytics.
Developer APIs and Integrations
The Agent API allows developers to use third-party foundation models while adding Perplexity tools including web search, URL retrieval, people search, finance search and an isolated coding sandbox. The Search API offers ranked results, domain controls, filtering and multi-query functionality. The Sonar API produces web-grounded responses with citations, conversation context and streaming. The Embeddings API converts text into vector representations for semantic search and retrieval-augmented generation. Perplexity also supports the OpenAI Chat Completions format, allowing developers to redirect compatible client libraries to Perplexity’s endpoint.
Current Perplexity Pricing Matrix
Public subscription pricing must be separated from API pricing. A Pro subscription does not automatically include unrestricted developer API calls. API consumption is billed independently on a pay-as-you-go basis.
Table 3: Perplexity AI 2026 Subscription Pricing & Query Limits
| Plan | Price | Pro Queries | Deep Research | Key Conditions |
| Free | $0 | Basic access | Limited | Lower access to premium features |
| Pro (monthly) | $20/month | 200/week | 20/month | 25 assets, 3 videos/month; files <50MB |
| Pro (annual) | $200/year | 200/week | 20/month | ~$17/month effective rate |
| Enterprise Pro (monthly) | $40/seat | ~400/week | ~50/month | ~2× Pro allowances |
| Enterprise Pro (annual) | $400/seat | ~400/week | ~50/month | ~$34/month effective rate |
| Enterprise Max (monthly) | $325/seat | ~4,000/week | ~500/month | ~20× Pro; 50-seat minimum for some features |
| Enterprise Max (annual) | $3,250/seat | ~4,000/week | ~500/month | ~$271/month effective rate |
| Large Enterprise | Negotiated | Custom | Custom | 250+ seat flexible pricing |
| API Platform | Pay-as-used | N/A | N/A | Separate billing; models, tokens, search calls |
API Pricing and Hidden Cost Drivers
Perplexity’s developer pricing is more complex than a single token rate. The Agent API passes through third-party model prices without a Perplexity markup; tool charges are added separately: web search at $0.005 per invocation, URL fetch at $0.0005, people search at $0.005, finance search at $0.005, sandbox at $0.03 per session, and search calls inside a sandbox at $0.005 each.
The standard Sonar model is listed at $1 per million input tokens and $1 per million output tokens, plus a request fee of $5 per 1,000 requests (low context), $8 (medium context) or $12 (high context). Sonar Deep Research uses a multi-part structure: $2 per million input tokens, $8 per million output tokens, $2 per million citation tokens, $3 per million reasoning tokens and $5 per 1,000 search requests. A long research job can cost more through reasoning, retrieval and citations than through the user’s initial prompt.
Rate Limits and Performance Bottlenecks
API accounts advance through usage tiers based on cumulative credit purchases. Tier 0 begins at no spend; Tier 1 requires $50 lifetime; Tier 2 requires $250; Tier 3 requires $500; Tier 4 requires $1,000; Tier 5 requires $5,000. Agent API capacity begins at 1 query per second and 50 requests per minute for Tier 0, rising to 33 requests per second at Tiers 4 and 5. The standalone Search API is documented at 50 requests per second with burst capacity of 50, independent of spending tier. Sonar Deep Research is limited to 5 requests per minute at Tier 0, rising to 20 at Tier 2.
Production deployments must account for search-request ceilings, burst traffic, Deep Research concurrency, third-party model latency, long context assembly, page-fetch failures, duplicate retrieval, citation processing, tool retries, connector authentication, external application quotas and cost amplification from agent loops. Developers should track cost per completed task rather than cost per initial prompt.
Technical Implementation Workflow
Step 1 — Select the Correct API: Use Search for raw ranked results with your own generation layer. Use Sonar for citation-backed answers. Use Agent for workflows requiring third-party models, tool execution or orchestration. Use Embeddings for semantic retrieval and similarity.
Step 2 — Create an API Key: Generate a key through the Perplexity API portal and store it in a secret manager, exposed via the PERPLEXITY_API_KEY environment variable rather than embedded in source code.
Step 3 — Install an SDK or Call REST: Python users can install the perplexityai package. TypeScript and direct HTTP requests are also supported. Teams using OpenAI-compatible clients can redirect compatible chat-completion calls to Perplexity.
Step 4 — Apply Retrieval Controls: Specify domains, search filters, recency requirements and context depth. Lower search context reduces request charges and latency. High context suits broad investigations where coverage matters more than speed.
Step 5 — Stream Long Responses: Streaming reduces perceived waiting time, especially for long answers. It does not reduce the total generation cost or retrieval workload.
Step 6 — Add Retry and Backoff Logic: Detect rate-limit responses, apply exponential backoff and avoid immediately repeating identical requests. Retries should be idempotent, particularly when agents take actions in external systems.
Step 7 — Log Every Cost Component: Store input tokens, output tokens, reasoning tokens, citation tokens, search calls, tool invocations and task duration. A single combined cost field hides the mechanism responsible for budget overruns.
Step 8 — Evaluate Answers and Citations: Citation presence does not guarantee that a source supports every sentence. Production systems should test entailment, freshness, source quality, duplication and coverage.
Step 9 — Add Human Approval for Actions: Research-only workflows can tolerate broader autonomy. Tasks that send messages, update customer records, make purchases or modify production systems should pause for approval.
Infrastructure Behind the Growth
In January 2026, Reuters reported that Perplexity signed a three-year, $750 million Microsoft Azure agreement. Microsoft confirmed that Perplexity selected Foundry as its primary AI platform, giving it access to models from several providers. Perplexity reportedly continued using Amazon Web Services as well.
“Foundry is Perplexity’s primary AI platform under the long-term arrangement.” — Microsoft spokesperson, Reuters, January 2026
The agreement does not prove a particular query count. It does show that Perplexity expects substantial and diverse inference demand. Multi-cloud and multi-model access can improve resilience, but it also introduces routing complexity. Efficient routing can reduce cost per answer even when raw query volume rises. Poor routing can make growth financially expensive despite subscription revenue.
Three Scenarios for the 2026 Query Run Rate
A credible analysis should use scenarios rather than presenting an unsupported point estimate.
Conservative Scenario
Monthly query volume grows 50 percent from the May 2025 level, reaching approximately 1.17 billion. This would indicate sharp deceleration from the reported 20 percent monthly rate but still represent major annual growth.
Expansion Scenario
Perplexity doubles or triples the May 2025 baseline, reaching approximately 1.56 billion to 2.34 billion monthly queries. This range is plausible if Comet, enterprise search and API distribution increased usage while consumer growth moderated.
Target-Achievement Scenario
Perplexity reaches the stated level of one billion queries per week, translating into about 4.35 billion monthly queries — more than a fivefold increase from May 2025. None of these scenarios should be labeled the official 2026 total. Until Perplexity releases a new number with a defined measurement period and query methodology, the correct editorial description is “undisclosed.”
Why Query Quality Matters More Than Raw Volume
Search businesses can inflate activity without creating corresponding value. Follow-up prompts, failed retrieval attempts, agent retries and fragmented tasks all raise query totals. The strongest metrics would include successful answers per query, retained users by cohort, searches per active user, paid conversion, cost per completed research task, citation accuracy, median and 95th-percentile latency, agent-task completion, human intervention rate, revenue per thousand queries and gross margin by product.
A 2026 study covering 24,000 searches across 243 countries found that AI search can reduce source diversity and concentrate exposure among fewer information providers. It also raised concerns about credibility and geographic variation. Volume alone cannot reveal whether an answer engine is increasing access to reliable knowledge or merely producing more synthesized text.
Expert Perspective on the Agentic Shift
“Resistance is futile regarding the spread of AI agents across connected devices.” — Cristiano Amon, Qualcomm CEO, Computex 2026
In an agentic environment, human prompts may grow slowly while machine-initiated searches grow quickly. The economically relevant unit may become the completed task rather than the individual query. Srinivas’s separate confirmation that Perplexity continues to plan for a 2028 public offering indicates that the company expects its operations to mature under investor scrutiny. An eventual public filing would likely require clearer definitions of users, queries, revenue concentration, compute commitments and gross margin.
Key Takeaways
- The latest strongly verified Perplexity AI monthly queries figure is 780 million for May 2025.
- That volume represents an annualized pace of 9.36 billion, not 52 billion. The 52 billion figure corresponds to the separate target of one billion queries per week.
- Perplexity has not publicly confirmed a precise companywide 2026 monthly-query total.
- Estimates of two billion to four billion monthly searches should be labeled projections unless supported by a primary disclosure.
- API, Comet and Computer workloads make a 2026 query different from a conventional search-box submission.
- Enterprise Max tier offers approximately 20 times the Pro query allowance and 25 times the Deep Research allowance at $325/seat/month.
- Businesses should evaluate completed tasks, retrieval accuracy, latency and cost rather than using raw query growth as the sole performance indicator.
Conclusion
Perplexity’s rise from an experimental answer engine to a multi-product research and agent platform is visible in its infrastructure, pricing and technical architecture. The 780 million-query benchmark from May 2025 remains significant: it represented an average of more than 25 million searches per day and confirmed that citation-based AI search had moved beyond a niche audience.
The 2026 picture is more complicated. Perplexity now processes consumer questions, enterprise knowledge requests, API calls, browser actions and long-running Computer tasks. Each can generate a different number of retrieval operations. A monthly total without a clear definition would therefore reveal less than it once did.
The responsible answer to “How many monthly queries does Perplexity handle in 2026?” is that no new official companywide figure has been publicly confirmed. A range of one billion to more than four billion may be discussed as scenario analysis, but it cannot replace a verified disclosure. Perplexity’s next meaningful milestone will not simply be a larger number — it will be a transparent explanation of what the company counts, how much each completed task costs and whether rising usage produces accurate, commercially sustainable results.
Frequently Asked Questions
How many monthly queries does Perplexity AI process in 2026?
Perplexity has not published a definitive companywide 2026 total. Its latest widely verified disclosure was 780 million queries in May 2025. Higher 2026 figures circulating online are estimates unless accompanied by a primary company statement.
Did Perplexity reach one billion queries per week?
Aravind Srinivas announced one billion weekly queries as a target for the end of 2025. No comparably authoritative public disclosure confirms that the target was achieved as of the writing of this article.
What is 780 million monthly queries annually?
Multiplying 780 million by 12 produces 9.36 billion queries per year. The often-cited 52 billion annual pace applies to one billion queries per week, not 780 million per month.
How many Pro searches can a Perplexity subscriber make?
The 2026 comparison page lists up to 200 Pro queries per week and 20 Deep Research queries per month for individual Pro subscribers at $20/month. Enterprise Max subscribers receive approximately 20 times the Pro search allowance and 25 times the Deep Research allowance.
Are Perplexity API requests included in its public query count?
Perplexity has not published a sufficiently detailed methodology explaining whether companywide query totals include every API request, internal agent search, URL fetch or consumer prompt. These categories should not be assumed to be equivalent.
References
Malik, A. (2025, June 5). Perplexity received 780 million queries last month, CEO says. TechCrunch. https://techcrunch.com/2025/06/05/perplexity-received-780-million-queries-last-month-ceo-says/
Perplexity AI. (2026). Perplexity API pricing. Perplexity Developer Documentation. https://docs.perplexity.ai/docs/pricing
Perplexity AI. (2026). Quickstart: Perplexity API. Perplexity Developer Documentation. https://docs.perplexity.ai/docs/getting-started
Perplexity AI. (2026). Rate limits and usage tiers. Perplexity Developer Documentation. https://docs.perplexity.ai/docs/rate-limits
Perplexity AI. (2026). Perplexity Enterprise pricing. https://www.perplexity.ai/hub/blog/perplexity-for-enterprise
Reuters. (2026, January 29). Perplexity signs $750 million AI cloud deal with Microsoft, Bloomberg News reports. https://www.reuters.com/technology/perplexity-signs-750-million-ai-cloud-deal-microsoft-bloomberg-news-reports-2026-01-29/
Aral, S., Li, H., & Zuo, R. (2026). The rise of AI search: Implications for information markets and human judgement at scale. arXiv. https://arxiv.org/abs/2506.01234
