TL;DR
Five tools, one decision rule: match the tool to your dominant task, not to its feature list.
- Sources you already have (PDFs, Docs, videos) → NotebookLM. Free, source-grounded, 50 sources per notebook.
- Fast cited web answers → Perplexity. Free tier is enough to start; Pro is $20/mo.
- A long written brief from the live web → ChatGPT, Gemini, or Claude Deep Research. Pick the one whose subscription you already pay for.
- Internal docs + live web in one answer → Claude Research (the only one that pulls Drive/Notion/Slack and the web together).
- One-claim fact check → Perplexity Quick Search.
If you only remember one number: at $19.99–$20/mo, Gemini Deep Research gives roughly 500 runs/month, ChatGPT Plus gives about 25/month (as of June 2026). For volume, that gap matters.
What each tool is
| Tool | Best for | Price (June 2026) | Citations | Reads your files |
|---|---|---|---|---|
| NotebookLM | Multi-doc synthesis on sources you upload | Free; Plus bundled in Google AI Pro $19.99/mo | Inline, back to your sources | Yes — 50 sources/notebook (free), 100 (Plus) |
| Perplexity | Fast cited web answers | Free; Pro $20/mo ($200/yr) | Inline, live web | Limited (Pro Spaces) |
| ChatGPT Deep Research | Long agentic web briefs | Plus $20 (~25 runs/mo); Pro $200 (much higher) | End-of-report | Yes, via attachments |
| Gemini Deep Research | Long briefs landing in Google Docs | Google AI Pro $19.99 (~500 runs/mo) | End-of-report | Yes, via Drive |
| Claude Research | Internal docs + live web together | Pro $20; Max $100 / $200 | Inline | Yes, via Connectors + uploads |
All figures are as of June 2026 and change often — re-check the vendor page before you subscribe.
NotebookLM
NotebookLM is Google’s source-grounded research notebook at notebooklm.google.com. You upload sources — PDFs, Google Docs, web links, YouTube URLs — and every answer is grounded in only those sources with inline citations, so the model cannot pull in random web pages on the side. It is free with any Google account.
Concrete limits on the free tier (June 2026): 50 sources per notebook, 100 notebooks, 50 chat queries per day, and 3 Audio Overviews per day. Each source can hold up to roughly 500,000 words. NotebookLM Plus — bundled with Google AI Pro at $19.99/mo — roughly doubles those caps (100 sources per notebook, far more daily chats).
Does best:
- Multi-document synthesis when you have 5–50 sources and want cross-cutting questions answered.
- Summarizing a long video lecture by pasting the YouTube link as a source.
- Audio Overviews — two AI hosts discuss your sources in a podcast-style summary you can listen to on a walk.
- Per-project knowledge bases you keep returning to.
Sample query:
Across these 12 PDFs of customer interviews,
list the 5 most-mentioned pain points,
with a direct quote for each.Not for: live web search, breaking news, or one-shot factoid lookup — it only knows your uploaded sources.
Perplexity
Perplexity is an AI search engine: ask a question, it runs a live web search and answers with inline source links you can click. The free tier is genuinely strong; Pro at $20/mo ($200/yr, ~$16.67/mo) unlocks Deep Research mode and effectively unlimited Pro Searches.
The main free-tier limit to know: free accounts get roughly 5 Pro Searches per day (the smarter, model-backed mode), with unlimited basic searches. If you hit that ceiling daily, Pro pays for itself.
Does best:
- Factual web queries where you need citations to skim and verify.
- “What happened today with X” — fresh news with sources.
- Comparison shopping where you want sourced links to specs and prices.
- Single-claim fact checks.
Sample query:
What changed in OpenAI's API pricing in the last 60 days?
List each change with a source link and a date.Not for: deep multi-doc analysis (no general upload library outside Pro Spaces), long structured reports, or grounding on your private docs.
ChatGPT Deep Research
ChatGPT Deep Research is an agentic research mode inside ChatGPT (Plus and Pro). You give it a prompt, it browses the web autonomously for roughly 5–30 minutes, then produces a long structured report with citations at the end. It is a separate product from ChatGPT’s regular search — slower, longer, more thorough.
The catch is volume. As of June 2026, ChatGPT Plus ($20/mo) gets about 25 Deep Research runs per month; the $200/mo Pro tier raises that ceiling dramatically. If you run more than one brief most days, Plus will run dry.
Does best:
- 15–30 minute deep dives where you want a written brief to come back to, not a chat thread.
- Market scans and competitor analysis.
- Due-diligence-style briefs that need 20+ sources synthesized into one document.
Sample query:
Produce a 2000-word brief on the indie AI-video-generator market:
vendors, pricing, model basis, recent funding,
with sources.Not for: quick lookups (use Perplexity), single-document analysis (use NotebookLM or Claude), or breaking news — by the time the run finishes, the freshest results are minutes stale.
Gemini Deep Research
Gemini Deep Research is Google’s equivalent inside Gemini (Google AI Pro or Ultra). It plans the research, browses the web, then writes a long structured Doc you can open directly in Google Docs.
Its standout trait at this price is headroom: as of June 2026, Google AI Pro at $19.99/mo allows on the order of 500 Deep Research runs per month — roughly 20x ChatGPT Plus’s ~25 at a near-identical price. Runs also tend to finish faster (often 2–10 minutes). ChatGPT’s reports are frequently more thorough, but if you run research daily, Gemini’s volume changes the math.
Does best:
- The same job shape as ChatGPT Deep Research, but favoring Google-indexed sources.
- High-frequency research, where ChatGPT Plus’s ~25/month cap would block you.
- Output that should land directly in Google Docs for editing and sharing.
Sample query:
Produce a 2000-word brief on the indie AI-video-generator market:
vendors, pricing, model basis, recent funding,
with sources.Not for: niche non-English topics where Google’s index is thinner, or grounding on private docs without first uploading them to Drive.
Claude Research
Claude Research is Anthropic’s equivalent inside Claude (Pro $20, Max $100/$200). It runs multi-step agentic web research and — uniquely among these five — can also pull from connected workspace tools via Claude Connectors: Google Drive, Notion, Slack, GitHub, and more.
Does best:
- Research that combines your internal docs and the live web in one answer — this is Claude’s real differentiation versus the others.
- Long-context analysis when you have also attached large PDFs (Claude models run a 1M-token context window as of June 2026).
- Careful citation behavior — it tends to surface where each claim came from.
Sample query:
Compare our Drive `Pricing v3` doc against current competitor pricing on the web;
flag every line where we are under or over,
with source links for the competitor numbers.Not for: pure quick lookups (use Perplexity); teams that live entirely in the Google stack and want output in Docs (use Gemini Deep Research).
Pick in 60 seconds
Name your dominant research task, then read across:
- I have the sources and want them cross-referenced (5–50 PDFs/Docs/videos) → NotebookLM. Free, best fit, source-grounded.
- I need fresh web facts with clickable citations → Perplexity. The free tier is enough to start.
- I want a long structured brief from the live web → ChatGPT or Gemini Deep Research. Pick the stack you already pay for; if you run research most days, weigh Gemini’s ~500/month against ChatGPT Plus’s ~25/month.
- I need internal docs + web in one answer, or large-PDF analysis → Claude Research.
- I just need to check one claim → Perplexity Quick Search.
Practical rule: try two of these on their free tiers for two weeks, then subscribe to the one that becomes muscle memory. Don’t buy three Deep Research subscriptions before you know which task you actually run weekly.
Common mistakes
- Treating the three “Deep Research” products as one thing. ChatGPT Deep Research, Gemini Deep Research, and Claude Research are three separate products on three separate subscriptions. A shared name does not mean a shared account.
- Buying for thoroughness, ignoring volume. ChatGPT’s reports are often the most thorough, but Plus caps you at ~25 runs/month. If you research daily, Gemini’s ~500/month at the same price may serve you better.
- Skipping NotebookLM because it is free. For “I have 20 PDFs and want them cross-referenced,” it is frequently the best fit, paid or not.
- Picking by feature list instead of by your real weekly task.
- Re-evaluating every month. Tools evolve, but your workflow should be sticky — switching costs more than the marginal feature gain.
FAQ
Q: NotebookLM or Perplexity for a literature review? A: NotebookLM if you already have the PDFs and want cross-references grounded only in those sources. Perplexity if you’re still discovering sources and need clickable citations from the open web. A common workflow uses Perplexity to find papers, then NotebookLM to digest them.
Q: Are ChatGPT Deep Research, Gemini Deep Research, and Claude Research the same thing? A: No — three separate products on three separate subscriptions. All run agentic multi-step searches, but each is bound to its own model and account, and their source pools, output destinations, and citation styles differ. One subscription does not cover another.
Q: Which Deep Research gives the most runs for the money? A: As of June 2026, Gemini Deep Research on Google AI Pro ($19.99/mo) allows roughly 500 runs per month, versus about 25/month on ChatGPT Plus ($20/mo) — close to 20x for a near-identical price. ChatGPT’s reports are often more thorough per run; choose by whether you value depth-per-run or volume.
Q: Is NotebookLM really worth picking over a paid tool? A: Yes, when the task is “I have 20 PDFs and need them cross-referenced.” Its source-grounded answers and Audio Overviews are best-in-class for closed-corpus work, and the free tier covers 50 sources per notebook across 100 notebooks.
Q: What makes Claude Research different? A: It’s the only one of the five that can pull your internal docs (Drive, Notion, Slack, GitHub via Connectors) and the live web into the same answer, with careful inline citations and a 1M-token context window for large attached PDFs.
Q: How often should I re-evaluate my choice? A: Quarterly at most. Re-evaluate when a major model release lands or when your task mix shifts — switching every month costs more than the marginal feature gain.
Related
Tags: #Tutorial #Research #Comparison