Perplexity has changed the default web retrieval mode in its Agent API: the low, medium and high presets now automatically use Fast Search. According to its October 2026 API changelog, web search pricing falls from $2.50 to $1 per 1,000 calls, with approximately 800 milliseconds less latency. The xhigh preset continues to use standard web search, and developers can explicitly set search_type: "web" to preserve the previous behavior.
This is an API retrieval change, not a confirmed update to the consumer Perplexity search ranking or citation system. Primary source: Perplexity API Changelog.
Which presets are affected?
| Preset | Default search mode |
|---|---|
| low | Fast Search |
| medium | Fast Search |
| high | Fast Search |
| xhigh | Standard web search |
The change affects applications that rely on defaults, potentially without a corresponding code deployment. Teams should record their actual preset and search configuration when comparing results over time.
Search calls cost 60% less
The new reported rate of $1 per 1,000 web search calls compares with $2.50 previously. That is a $1.50 reduction per thousand, or 60%. At 100,000 eligible calls, the difference would be $100 instead of $250, excluding other model, token or tool charges. This is an illustrative calculation, not a total API invoice.
Approximately 800 milliseconds faster
Perplexity reports about 800 ms lower search latency. The improvement is provider-reported and may not translate into the same reduction for complete agent tasks. End-to-end speed also depends on model inference, orchestration, network behavior and the number of tool calls.
Keeping the standard search behavior
Developers who want the previous search mode can use the documented setting:
"search_type": "web"This is a configuration fragment rather than a complete runnable API request. Developers should confirm its exact placement and compatibility in the current Agent API reference.
Will Fast Search change website citations?
The changelog does not establish any specific effect on citations. A different retrieval mode is a reason to test source selection, but neither lower cost nor lower latency proves that websites will be cited more often, less often or differently. The consumer search product must also be distinguished from Agent API behavior.
A controlled comparison would keep prompts, model settings and other variables stable while testing Fast Search against standard search. Researchers should compare cited URLs, source overlap, factual quality and variation across repeated requests.
Why this matters for GEO and AI visibility
Visibility studies that label all observations simply as 'Perplexity' can hide important differences in product surface, model, preset and retrieval mode. Researchers should preserve the test date, model, preset, search_type, prompt and cited sources. Otherwise, a shift in the retrieval default could be mistaken for a change in brand visibility.
Practical checklist for developers
- Audit requests using low, medium or high presets without an explicit search type.
- Compare representative queries under both retrieval modes where supported.
- Measure response quality, citations and latency separately.
- Calculate full workload costs, including tokens and other tools.
- Set
search_type: "web"when standard retrieval must remain consistent. - Document any change to defaults in longitudinal AI visibility research.
What the announcement does not prove
It does not establish that Fast Search is less accurate, that standard search always provides better sources, that SEO ranking factors changed or that the consumer Perplexity interface uses the same new defaults.
FAQ
Which presets now use Fast Search?
Low, medium and high.
Does xhigh change?
No. It continues to use standard web search.
What is the new web search price?
$1 per 1,000 calls, compared with $2.50 previously.
How can developers force standard search?
Set search_type: "web" in the appropriate API configuration.
Are AI citations affected?
The announcement does not confirm any citation impact; controlled tests are needed.
NetContentSEO analysis
Perplexity's Fast Search rollout is a meaningful change in agent economics and experimental reproducibility. It lowers the reported cost and latency of retrieval while adding a configuration variable that GEO researchers must record. The next step is evidence-based comparison of retrieval quality and cited sources, not assuming that a faster search mode changes organic visibility.
Source: Perplexity API Changelog, October 2026.