Industry News
Perplexity retires its Sonar API today: what changes for brand citations
Perplexity's Sonar API retired on September 27, 2026. Here's what the shift to Agent API means for brand citation behavior in AI search — and what to fix now.
On this page
# Perplexity retires its Sonar API today: what changes for brand citations
By Stephan Charles | Last fact-checked: September 27, 2026
BrandCited is a GEO (Generative Engine Optimization) platform that tracks brand visibility across 9 AI search engines: ChatGPT, Perplexity, Claude, Gemini, Copilot, Grok, DeepSeek, You.com, and Brave. It assigns each brand an AI Visibility Score from 0 to 100 and runs 30+ audit checks to surface exactly what needs to change.
On September 27, 2026, Perplexity shut down its Sonar API — the endpoint that powered hundreds of marketing tools, SEO dashboards, and AI search monitoring platforms since 2024. The replacement is Perplexity's Agent API, which launched August 13, 2026. For developers, this is a migration. For brands, it is a forcing function: understand why Perplexity's citation behavior differs from every other AI engine, or keep losing the citations that Perplexity generates at 22 times the rate of ChatGPT. The transition also lands as HP prepares to pre-install Perplexity on Windows PCs starting October 2026 — putting Perplexity a single taskbar click away from every HP business user before end of year.
Run a free AI visibility audit at brandcited.ai to see your current Perplexity citation score across all 9 engines before reviewing the gaps below.
What did Perplexity announce on August 13, 2026?
Perplexity launched the Agent API as a unified surface for LLM inference, live web search, and agent orchestration — and set a hard deadline: September 27, 2026, Sonar endpoints retire. The official announcement states: "On September 27, 2026, the Agent API becomes the main surface for this work, and Sonar tiers retire on the same date."
The migration path splits by model ID. The base perplexity/sonar model moved to the Agent API on September 25 with no changes to request parameters, citation format, or pricing. Tools that called sonar-pro or sonar-reasoning-pro face a harder cut: Perplexity's Agent API rejects both model IDs outright, and no direct equivalent exists in the new API surface.
The Agent API is structurally different from the old chat-completions endpoint. It supports multi-step agentic workflows, richer retrieval controls — domain allow/deny lists of up to 20 domains per request, date range filters, regional search targeting by ISO country code — and a search_results response object that returns each cited URL with its title, domain, publish date, and relevance score. That search_results object is what matters for brands: citation data from Perplexity is now machine-readable at the platform level for the first time.
Why does Perplexity's citation rate matter more than ChatGPT's?
Perplexity cites brands in 13.05% of responses. ChatGPT cites brands in 0.59% of responses. A 2026 analysis of 34,234 AI responses by QuickSEO puts the gap at 22 times in citation frequency between the two platforms.
That gap matters most for high-consideration queries. Perplexity users run longer, more specific searches: "best project management software for engineering teams under 50 people" rather than "project management software." A brand that earns a citation on that query sits inside the buyer's research workflow before the buyer has visited a single vendor site.
Only 11% of domains appear in both ChatGPT and Perplexity citation pools for the same queries, according to a 2026 platform overlap audit by AuthorityTech. A brand that appears in ChatGPT answers and assumes Perplexity follows is leaving 89% of Perplexity citations open to competitors. The two engines retrieve from different source pools and weight different signals.
Perplexity processes hundreds of millions of queries per month. In October 2026, HP will begin pre-installing Perplexity on Windows PCs, starting with the HP ZBook Ultra G3a and expanding to more Windows devices. That distribution deal grows Perplexity's query volume from business users before end of year.
Run a [free BrandCited scan](https://www.brandcited.ai) to see your current citation rate on Perplexity and 8 other AI engines.
What does the Agent API change for how Perplexity chooses which brands to cite?
The Agent API adds richer query context to Perplexity's retrieval layer, and that changes which content surfaces. Under the Agent API, Perplexity can run multi-step research tasks: first query a topic, retrieve citations, then run follow-up queries on sub-topics inside a single agentic session. The engine cites sources across all steps, not only in the final answer.
Two changes matter most for content strategy. First, domain credentialing is now explicit at the API level. Applications can supply allow/deny domain lists per query, meaning enterprise tools built on the Agent API constrain which brand domains are eligible for citation before retrieval ranking runs. Brands not indexed by Perplexity's crawler, or blocked by restrictive robots.txt directives, get excluded at the infrastructure level.
Second, the Agent API's structured citation response returns search_results objects with explicit date metadata. Content freshness carries a structural weight in the response format. Pages missing machine-readable publish dates — specifically datePublished in Article schema and date markup in body copy — rank below dated equivalents under the Agent API's retrieval logic. LLM Gateway documented this change in its Sonar retirement brief.
What breaks for brands whose monitoring tools were built on Sonar?
Any tool that issued requests to sonar-pro or sonar-reasoning-pro stopped returning Perplexity data at midnight on September 27, 2026. The symptom in dashboards is not an error message. It is silence: citation tracking shows zero new citations, Perplexity rows go stale, and share-of-voice charts flatline without explanation.
Two categories of tools are affected. AI search monitoring platforms that tracked brand mention rates used sonar-pro for its higher-quality retrieval. Brand intelligence tools that pulled Perplexity for competitive citation analysis used sonar-reasoning-pro for its reasoning-augmented responses. Both model IDs are dead as of today.
BrandCited's Perplexity monitoring connects through the Agent API. If your brand monitoring dashboard shows a Perplexity data gap starting today, the issue is at your tool provider's integration layer — not within BrandCited's monitoring pipeline.
What should brands do right now?
- Check your Perplexity citation rate today. Run a free BrandCited scan at brandcited.ai. The scan queries Perplexity with your brand name, your product category, and your top three competitor comparison queries, then returns a citation count and coverage rate. A rate below 5% means structural content gaps that the Agent API era widens.
- Verify that Perplexity can crawl your site. Check your
robots.txtfor directives blockingPerplexityBot. The Agent API's application-layer domain allow/deny lists cannot override a crawl block. If Perplexity has not indexed your pages, no API query surfaces your brand. - Add `datePublished` and `dateModified` to every Article schema block. Perplexity's structured citation response tracks content freshness. Pages without machine-readable publish dates rank below dated equivalents in Agent API retrieval. The Schema.org Article type documents both fields.
- Add FAQPage schema to your key category pages. Perplexity's Agent API handles multi-step queries where follow-up questions are common. Pages with FAQPage schema give Perplexity structured answer candidates for those follow-up steps. Pages without schema require full-text retrieval parsing, which is less reliable.
- Add structured author markup. Perplexity weights author credibility as a citation signal. An
authorfield in Article schema with asameAslink to the author's LinkedIn profile increases the credibility score Perplexity assigns to the page before citation ranking runs.
BrandCited's audit engine checks Perplexity crawlability, Article schema completeness, datePublished freshness, FAQPage schema presence, and author markup as part of its 30+ audit checks. Run a free scan at [brandcited.ai](https://www.brandcited.ai) to see which gaps apply to your brand across Perplexity and 8 other engines.
AI search updates from the last 24 hours
- Perplexity + HP partnership: HP will pre-install Perplexity on Windows PCs starting with the ZBook Ultra G3a in October 2026, with more Windows devices to follow. The deal puts Perplexity's sourced search inside every HP business user's taskbar.
- OpenAI GPT-6 Cyber preview: OpenAI previewed GPT-6 Cyber at DevDay on September 24 alongside a new enterprise security product. GPT-6 Cyber is scoped to cybersecurity workflows and does not change ChatGPT's general brand citation behavior.
- 5WPR Citation Source Index 2026: 5W Research released the AI Platform Citation Source Index 2026, naming the 50 websites that now determine brand visibility inside ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews.
- Google AI Overviews CTR data: Seer Interactive's data shows organic CTR dropped 61% for queries that trigger AI Overviews — from 1.76% to 0.61% — while brands cited inside Overviews earn 35% more organic clicks than brands that rank but go uncited.
How does BrandCited audit Perplexity visibility?
BrandCited queries Perplexity through the Agent API for every brand in its monitoring queue. The audit checks crawlability, Article schema completeness, datePublished freshness, FAQPage schema presence, and author credibility markup. Each failing check surfaces as a ranked finding with the specific schema field or robots.txt line to change. The scan runs in 30 seconds and is free at brandcited.ai.
Frequently asked questions
What is Perplexity's Agent API and how is it different from Sonar?
The Agent API is Perplexity's unified interface for LLM inference, web search, and multi-step agent workflows. It replaced the Sonar API on September 27, 2026. The Agent API returns structured search_results objects with domain, title, date, and relevance data alongside citations, making citation data machine-readable. The Sonar API returned unstructured text with inline citation markers only.
Does Perplexity's Sonar API retirement affect brands not using the API directly?
Yes, indirectly. Many third-party SEO and AI monitoring tools used the Sonar API to track Perplexity citations on behalf of customers. Tools that called sonar-pro or sonar-reasoning-pro stopped returning Perplexity data on September 27. If your brand monitoring dashboard shows a Perplexity data gap starting today, the issue is at your tool provider's API integration layer.
Which Perplexity model IDs were discontinued?
sonar-pro and sonar-reasoning-pro were discontinued on September 27, 2026. The base sonar model migrated to the Agent API with no changes to request parameters or citation format. Tools that called sonar-pro or sonar-reasoning-pro now receive errors from Perplexity's API.
What citation rate should my brand target on Perplexity?
Perplexity cites brands in 13.05% of responses on average across all query types, according to a 2026 study of 34,234 AI responses. Category leaders in B2B SaaS score above 20% on their primary product category queries. A rate below 5% indicates structural content gaps.
How often should brands check their Perplexity citation rate?
Weekly. Perplexity's retrieval layer updates with new web content in near real-time. A competitor that publishes a strong comparison page or earns coverage on a high-authority domain can displace your brand's citations within days. BrandCited monitors Perplexity weekly and alerts when citation rate drops more than 10 percentage points week-over-week.

