If you have ever typed a question into an AI search tool and watched a little spinner turn for what felt like forever, you already understand why speed matters. Perplexity AI search just made a significant move to fix that problem, and the numbers behind it are worth understanding even if you have never written a line of code.
Perplexity, the AI-powered search company, released a new internal engine called Photon in late September 2026. According to Perplexity, this engine cut a key speed measurement called p99 latency from about 800 milliseconds down to about 65 milliseconds. P99 latency is a way of measuring how long the slowest one percent of searches take. Think of it as the worst-case wait time for real users on a busy day. Cutting that number from 800 ms to 65 ms means the bad days got a lot better.
What Perplexity AI Search Actually Changed Under the Hood
Photon is a retrieval and ranking engine, meaning it is the part of the system that finds relevant documents and decides which ones to show you first. Perplexity built it from scratch in a programming language called Rust, which is known for speed and reliability. It replaced an older open-source engine the company had been modifying for its own needs.
The old engine had three growing problems as Perplexity’s index of web content expanded. First, slow searches were getting slower, with that worst-case wait time sitting near 800 milliseconds. Second, when the system merged new data into the index, wait times spiked to about 1.2 seconds for stretches of 10 to 15 minutes, according to Perplexity. Third, if something went wrong and a backup cluster needed to be brought online, the process could take more than a week.
Photon addresses all three. According to Perplexity, it also runs on about 20 percent fewer serving machines than the old system and stores about 2.5 times as much data per document, which the company used to improve ranking quality.
What This Means in Everyday Life
You do not need to understand how Photon works to benefit from it. What matters is that faster, more reliable search results affect real tasks people do every day.
A small business owner researching suppliers
- Problem: Comparing five potential vendors takes multiple searches, and slow results break concentration and waste time.
- Usual approach: Open several browser tabs, wait for each search to load, and manually piece together findings.
- With AI: Use Perplexity to ask a specific question like: find me recent reviews of wholesale packaging suppliers in the United States, summarize the top options, note any complaints about shipping times, and keep the answer under 200 words.
- What to provide: The task (find and summarize), the context (wholesale packaging, United States), the format (summary), the audience (a business owner making a purchase decision), and the constraint (brief answer).
- Benefit: Faster retrieval means the answer arrives more quickly, and the new Fast Search mode costs $1 per 1,000 requests, according to Perplexity, making it affordable for regular use.
- Limit: AI search can miss recent or niche information. Always verify supplier details directly before committing money.
An employee preparing a briefing document
- Problem: A manager needs a quick overview of a competitor’s recent announcements before a meeting in an hour.
- Usual approach: Search manually, skim several articles, and type up notes under time pressure.
- With AI: Ask Perplexity to summarize recent public news about a specific company, formatted as three bullet points, written for a non-technical audience.
- What to provide: The task (summarize), the context (specific company name, recent news), the format (three bullets), and the audience (non-technical manager).
- Benefit: A faster engine means less waiting during a time-sensitive task, and the structured prompt produces output that is easier to review quickly.
- Limit: AI search pulls from publicly available sources. It will not surface internal documents or paywalled reports. Human review before the meeting is essential.
A freelancer tracking industry trends
- Problem: Staying current across several topics takes time that cuts into billable hours.
- Usual approach: Subscribe to newsletters and skim them when time allows, often falling behind.
- With AI: Run a weekly Perplexity search asking for the three most discussed developments in a specific field over the past seven days, with a one-sentence explanation of why each matters to small businesses.
- What to provide: The task (identify and explain), the context (specific field, past seven days), the format (three items with one sentence each), and the audience (small business context).
- Benefit: Consistent, fast results make a weekly habit easier to maintain.
- Limit: AI search reflects what is publicly indexed. Emerging stories that have not been widely covered yet may not appear.
The Cost of Waiting on Better AI Search Tools
Perplexity also introduced a Fast Search mode that pairs Photon with lighter ranking designed for automated workflows. According to Perplexity, across 3,554 tasks on six benchmarks, Fast Search scored 64.3 percent accuracy at an estimated cost of $59.73, compared to 64.0 percent accuracy for the default mode at $187.60. That is roughly 68 percent cheaper for comparable task quality, according to Perplexity.
The trade-off is real and worth knowing. On Perplexity’s internal long-tail benchmarks, a relevance measure called DCG dropped from 2.45 to 2.21 with Fast Search, and answer availability fell by 2.9 percentage points. Perplexity recommends Fast Search for routine, day-to-day tasks and the default mode for harder or more ambiguous questions.
For individuals and small businesses, the practical message is straightforward. Faster and cheaper AI search tools lower the barrier to using them regularly. The cost of not experimenting is that competitors and colleagues who do adopt these tools build habits and skills that compound over time. That gap is documented in how businesses that integrate AI tools earlier tend to find workflow improvements sooner, as explored in 10 Powerful Benefits of AI in the Workplace.
How to Start Using Perplexity AI Search Today
Perplexity is available as a web tool at perplexity.ai. You do not need a paid account to try basic searches. The Fast Search mode through the API, priced at $1 per 1,000 requests according to Perplexity, is aimed at developers building automated tools, but the standard search interface is accessible to anyone.
The most useful first step is replacing one regular search habit with a Perplexity search this week. Pick a task you already do, such as researching a product, summarizing a topic for a colleague, or checking on industry news. Notice whether the results save you time compared to a standard search engine.
If you are thinking about how AI tools fit into a broader workflow at your business, it helps to start with a specific problem rather than a tool. The Problem-First Implementation Framework is a useful way to think through that before spending money on subscriptions.
Limitations and Responsible Use
Photon and Fast Search are vendor-reported results. Perplexity’s own benchmarks show the trade-offs honestly, but independent comparisons under identical conditions are not available from the source. Treat the numbers as directionally useful, not as guarantees.
For privacy, avoid entering confidential business information, client names, or sensitive personal details into any AI search tool unless you have read and understood the platform’s data handling policies. AI search tools pull from public web indexes and do not have access to private or proprietary information unless you provide it directly.
Accuracy is never guaranteed. AI search can confidently return outdated or incomplete information. Any result that will influence a business decision, a legal matter, or a financial choice needs to be verified through primary sources. Human review is not optional.
For teams thinking about AI adoption more broadly, professional AI training helps people develop the judgment to know when to trust AI output and when to dig deeper. That skill matters more as these tools get faster and easier to use.
Speed Is a Feature, Not Just a Spec
A drop from 800 milliseconds to 65 milliseconds in worst-case search latency is not just a technical achievement. It changes the experience of using AI search from something that sometimes feels sluggish to something that feels responsive. For people who are still on the fence about whether AI tools are worth the effort to learn, that friction reduction matters. Tools that feel fast and reliable are tools people actually use. And tools people actually use are the ones that deliver real benefits over time.
Your first step
Open perplexity.ai and type a question you would normally search on Google, such as asking for a plain-language summary of recent news in your industry. Notice how the results compare in speed and usefulness to your usual search habit. Try rephrasing the question to include what format you want and who the answer is for, then compare the two results.