Perplexity is the research tool I open 30 times a day. Most founders use it like a faster Google. That’s leaving 70% of the value on the table. I run 500k.io solo. My research bill before AI was about $400/mo across SEMrush, news subscriptions, and a part-time research VA. Today it’s $20 — Perplexity Pro — and the work is faster and more thorough than the old stack. After 9 months daily-driving it across 4,000+ research sessions, this is the workflow.
Quick context: I run The Kreators AI — the agency I co-founded with Jack — where we manage about $45M in client revenue ($10M Meta personally, $35M on Jack’s side). Solo on 500k.io I’m at $9,500 MRR with one Meta Ads client = $114K ARR, 22.8% of the way to the $500K target. Research compounds. Bad research compounds badly. Perplexity is the tool that keeps mine clean.
What is Perplexity in one paragraph?
Perplexity is an answer engine that runs a real-time web search, synthesizes the top sources, and returns an answer with inline citations. Each citation is clickable. You can ask follow-ups in the same thread and the model retains context. Pro Search ($20/mo) lets you choose models (GPT-5, Claude 4.6 Sonnet, Gemini 2.5, Grok 3, or their default). Deep Research mode runs a multi-step agentic search and outputs a 10-30 page report. Spaces let you bundle related searches into a project folder.
That’s the surface. The 7 workflows below are where the leverage actually lives.
The 7 Perplexity workflows I run weekly
1. SERP scan before I write anything
Before I write an article on 500k.io, I ask Perplexity: “What are the top 10 articles ranking for ‘how to write a Claude.md file’ as of May 2026? For each, give me the title, URL, and a 2-sentence summary of their angle.”
This used to take me 45 minutes in SEMrush + manual reading. Now: 90 seconds. The citations let me click into the actual articles when one looks interesting.
The follow-up I always run: “What angles are missing from these 10 articles? What questions do they fail to answer?”
That’s where I find the opening.
2. Competitor deep-dive (Deep Research mode)
For competitor research I switch to Deep Research mode. The prompt template:
“Run a deep analysis of [competitor name]. Cover: their pricing, their last 6 months of product launches, their content cadence, their reported revenue/funding, the top 3 weaknesses in their offering, and 5 testimonials or reviews from real customers.”
Deep Research takes 8-12 minutes and outputs a structured report. I save it as PDF, drop it in my Notion competitor folder. The reports aren’t perfect — they hallucinate occasionally on private financial data — but they’re 80% of what a $2,000 freelance analyst would deliver.
3. ICP research (the underrated workflow)
When I’m trying to understand a buyer profile, Perplexity beats LinkedIn Sales Navigator for the qualitative work. Prompt:
“What are the top 5 frustrations of solo Meta Ads buyers spending $5K-$20K/month? Cite from Reddit threads, Twitter, and forums published in the last 6 months.”
The Reddit citation source is gold. Perplexity surfaces real complaints in real customer voice. I copy the language directly into landing page copy.
4. News monitoring (Spaces)
I have a Perplexity Space called “AI tools — daily.” I drop one prompt into it each morning:
“What are the most important AI tool launches and pricing changes from the last 24 hours? Focus on Claude, OpenAI, Google AI Studio, Cursor, Windsurf, and Perplexity itself. Skip noise.”
The Space remembers the context. I can ask follow-ups across days. This replaced my $20/mo subscription to a curated AI newsletter — the newsletter was usually 12 hours behind Perplexity anyway.
5. Statistic verification
The cheapest way to ruin an article is publishing a stat that’s wrong or sourced from a SEO blog regurgitating another SEO blog. Before I publish any number, I run:
“Verify this claim: ‘[claim]’. Find the original primary source. If you can’t find a primary source, mark it unverified.”
Perplexity will tell you when a stat is uncited. That alone is worth $20/mo.
6. Expert opinion synthesis
When I want to understand what real practitioners think about a topic — not what AI thinks — I ask:
“What do experienced [role] say about [topic] in 2026? Cite from podcasts, conference talks, and long-form posts. Give me 5 distinct positions.”
This is how I built my opinion on AI agent monetization. I read a Latent Space podcast transcript about agent economics, a swyx essay, and a Stratechery piece — all surfaced by Perplexity in 4 minutes.
7. Pre-writing for articles (the meta workflow)
This article you’re reading? It started with one Perplexity prompt: “What are the top questions solopreneurs ask about Perplexity that aren’t well-answered by current SERP results?” The answer surfaced 6 angles. I picked 7. The H2 structure of this article is downstream of that single Perplexity query.
Pro Search vs Deep Research — when to use which
| Mode | Time | Use for | Cost |
|---|---|---|---|
| Quick Search | 5-10s | Definitions, single facts | Free |
| Pro Search | 15-30s | 90% of research questions | $20/mo |
| Deep Research | 5-15min | Competitor reports, market sizing, full ICPs | $20/mo (limited daily uses) |
| Labs | Variable | Experimental tools (image gen, page builder) | $20/mo |
Default to Pro Search. Move to Deep Research when the answer needs to look like a report you’d send to a client.
The features most people skip
Spaces
Spaces are project folders. Each Space has its own context — uploaded files, custom instructions, persistent thread history. I have one for each major topic: “AI tools daily”, “Solopreneur ICP”, “GEO research”, “Bathroom remodeling US (UpgradeMatch)”.
Most users never create a single Space. That’s the equivalent of using Notion without pages.
File upload
Perplexity reads PDFs, Word docs, spreadsheets. I upload my client’s last 90 days of Meta Ads exports and ask “what creatives are showing CTR decay?” The model parses the CSV and answers in 30 seconds.
Focus filters
Under the search bar there’s a “Focus” dropdown: Web, Academic, Social, Reddit, YouTube. For ICP research I default to Reddit. For citation-quality work I default to Academic. Most people leave it on Web.
Threads
Every search creates a thread. Threads are shareable URLs. When Anastasia asks me a question I’ve already researched, I send her the Perplexity thread instead of re-explaining. Free tier supports this.
Why I cancelled ChatGPT Pro because of Perplexity
“ChatGPT Search is fine for one-off lookups. Perplexity is built for the founder who spends 60+ minutes a day in research. They’re not the same product. They just look the same in screenshots.”
Here’s the honest math. I was paying:
- ChatGPT Pro: $200/mo
- Perplexity Pro: $20/mo
- Claude Pro: $20/mo
Total: $240/mo. Three subscriptions doing partially-overlapping things.
After auditing, the breakdown:
- 70% of my AI work happens in Claude (writing, code, reasoning).
- 25% happens in Perplexity (research, citations, news).
- 5% happens in ChatGPT (occasional GPT-5 specific tasks).
5% doesn’t justify $200/mo. I cancelled ChatGPT Pro. Now: Claude Max $100 + Perplexity Pro $20 = $120/mo. Same output. -$120/mo.
The full breakdown of my 13-tool stack at $565/mo lives on the stack page.
What Perplexity is bad at
Three things, honestly.
Local results. Perplexity is brutal at “find me a coffee shop in Lisbon with wifi”. Google still wins.
Buying things. Perplexity has shopping features but they’re shallow. For real product research with reviews + price comparisons, I still hit Google + Reddit manually.
Real-time creative work. It’s a research tool. If I want to draft an ad hook, I open Claude. If I want to evaluate the hook against my ICP’s pain points, I open Perplexity to verify the pains are real.
The 90-second daily routine
Every morning, before I open email, I run 3 Perplexity queries:
- “Top AI tool launches in the last 24 hours, focus on solopreneur-relevant tools.”
- “Top developments in [my current project’s vertical] in the last 24 hours.”
- “Any pricing changes or platform updates from Beehiiv, Cloudflare, Anthropic, Stripe today?”
Total time: 90 seconds. I read the answers, scan the citations for anything worth deeper. By 8am I’m caught up on the AI ops that matter to me. This replaced about 6 newsletters I used to skim.
Internal links to dig deeper
- The honest math: $500K solo SaaS in 18 months — what research compounds into.
- The $0 stack: 12 free tools for solopreneurs — Perplexity at the free tier.
- Claude Code: a $500K founder’s first 30 days — research feeds into the content factory.
- GEO 2026: how to get cited by ChatGPT and Perplexity — the meta loop: be on the receiving end of Perplexity citations.
- MCP explained for founders — pair Perplexity with Bright Data MCP for compound effect.
- Notion vs Obsidian knowledge stack for AI builders — where Perplexity outputs go to live.
External sources
- Perplexity official documentation — auth-free reference for current features.
- Latent Space — Aravind Srinivas interview — Perplexity’s founder on the architecture.
- Stratechery — answer engines vs search engines — Ben Thompson on what Perplexity changes.
What’s next
Open Perplexity. Set a single Space called “Daily ops.” Add three prompts. Run them tomorrow morning before email. That’s the entire onboarding.
If you don’t notice a difference in 7 days, cancel. I’d refund anyone who reports back that it didn’t move their workday. So far: zero refunds.
FAQ
Is Perplexity worth $20/mo when ChatGPT and Claude have web search?
For research workflows, yes. Perplexity cites sources inline, supports follow-up threads with persistent context, and its Spaces feature behaves like a research project folder. ChatGPT Search is fine for one-off lookups; Perplexity wins when you're spending 60+ minutes on a topic.
Does Perplexity replace Google Search?
For research-heavy queries, yes. For commercial intent (buying things, finding stores, local results), no. I still hit Google for product reviews and Maps. Perplexity replaces about 40% of my Google sessions.
What's the biggest mistake new Perplexity users make?
Treating it like a chatbot. Perplexity is a citation engine — its value is the trail of sources under each answer. If you're not clicking into the citations, you're paying $20/mo for a worse ChatGPT.
Should I use Pro Search or Deep Research mode?
Pro Search for under-30-second questions. Deep Research for the things you'd otherwise hire an analyst for. Deep Research takes 5-15 minutes but produces a structured 10-30 page report with citations. Use it for competitor analysis, market sizing, vertical-specific deep-dives.
Can I use Perplexity for SEO research?
Yes — and I'd argue it's the best AI tool for SERP research. Ask it 'what are the top 10 articles ranking for X' and it'll cite them with snippets. Pair with a Bright Data MCP and you've replaced a $99/mo SEO tool for most use cases.
Why does Perplexity sometimes get the date wrong?
Because it weights recency loosely. If a stat is time-sensitive (e.g., 'Claude pricing 2026'), explicitly state the year and ask Perplexity to filter for sources from the last 3 months. The 'Last 3 months' filter in the UI works.