45 Best Prompts for Gemini AI (2026): Search, Research, Images, Video

Google's Gemini app passed 750 million monthly users at the end of 2025, according to Google's Q4 earnings reported by TechCrunch, up from 450 million earlier in the year. Yet most "best Gemini prompts" lists could be "best ChatGPT prompts" with the names swapped. They ignore the whole reason to open Gemini in the first place: it does several things the other tools can't.

This is a collection of 45 ready-to-use Gemini prompts, organized by the thing each one takes advantage of: live Google Search, Deep Research, long-document analysis, multimodal input, image, video, and music generation, Workspace, and code. Find the one that fits what you're doing, fill in the brackets, and refine the result until it sounds right. If you want to learn the underlying skill, our guide to writing prompts covers the how; this page hands you the finished ones.

Gemini is free at gemini.google.com, running Gemini 3.6 Flash by default as of late July 2026, with Deep Research, Canvas, and Gems included and limited access to the top model, Gemini 3.1 Pro. Image generation uses its Nano Banana model, and Google added video (Veo) and music (Lyria) generation during 2026, though those mostly need a paid tier. Paid plans add the newest models and higher limits, but the free version handles almost everything below.

A good prompt plays to Gemini's strengths

A good Gemini prompt does one thing a generic prompt doesn't: it points Gemini at a capability the other tools lack. Ask it to research like a chatbot working from memory and you get a flat answer:

Research electric cars for me.

Point it at live search with a clear shape, and you get something current and checkable:

Use Google Search to research the three best-selling electric cars under $40,000 in 2026. For each, give the range, the starting price, one standout feature, and a recent review source with a link. Put it in a table, and flag anything where the data looks unclear.

That's the whole idea behind this list. A prompt that could run on any chatbot wastes what makes Gemini worth opening in the first place. The strongest ones tell it which capability to reach for, whether that's searching the live web, reading a long PDF, looking at an image, or generating one, and then hand it the specifics to work with. Match the prompt to the capability and the gap between Gemini and a flat, generic answer gets obvious fast.

Each of the 45 prompts below is built to use something Gemini is genuinely good at, with a [bracket] or two for the details only you can supply. Fill those in, attach any files that help, and treat the first answer as a draft to check, not a final word.

What Gemini can do that a plain chatbot can't

Gemini's advantage is a handful of capabilities most chatbots don't have, and the strongest prompts each target one of them. This map shows what to reach for, and what to point it at.

Research with live Google Search

Gemini's signature edge: it can pull current information from Google Search mid-answer, so these point to real, recent sources you can open.

1. Find current, credible sources

The fix for "I can't find anything solid on this."

Use Google Search to find five recent, credible sources on [topic]. For each, give the title, the publisher, the year, a one-line summary, and the link. Prefer primary sources, journals, and official data over blogs.

2. Fact-check a claim

Settle whether something is actually true.

Fact-check this claim against current sources using Google Search: '[claim]'. Tell me whether it holds up, what the evidence says, and link the strongest source. Flag anything that's outdated.

3. Get a current market snapshot

A quick read on what's happening now.

Using Google Search, give me a current snapshot of [company, product, or market]: what's changed in the last few months, the main players, and three recent sources. Note where the information is thin or conflicting.

4. Compare options with live data

Two choices, weighed on current facts.

Compare [option A] and [option B] using current information from Google Search: price, key features, pros and cons, and what recent reviews say. Put it in a table and cite a source for each row.

5. Brief yourself before a meeting

Walk in knowing who you're talking to.

I have a meeting about [topic] with [person or company]. Use Google Search to brief me: who they are, any recent news, and three smart questions I could ask. Keep it to one page.

Deep Research reports

Deep Research is Gemini's agentic mode: it browses many sources on its own and writes a cited, multi-page report. Use it when one answer isn't enough. It takes a few minutes rather than a few seconds, and the payoff is a document you could hand to someone, with a trail showing where each claim came from.

6. Run a full research report

A structured report instead of a paragraph.

Use Deep Research to produce a structured report on [topic]. Cover the background, the current state, the key players or data, the opposing views, and what's likely next. Cite sources throughout, and flag anything uncertain.

7. Review the thinking on a question

The state of play on a debated topic.

Use Deep Research to review the current thinking on [question]. Summarise the main positions, the strongest evidence for each, where experts disagree, and the gaps still open. Include a source list.

8. Analyze a market or industry

The numbers and the shape of a sector.

Run a Deep Research report on the [industry] market: size and growth, the major companies, recent shifts, and the main risks. Use current data, and cite each figure with its source and year.

9. Deep-dive a buying decision

A thorough compare before you commit.

Use Deep Research to help me choose between [options] for [use case]. Compare them on the criteria that matter, summarise real user feedback, and end with which one fits which situation. Cite your sources.

10. Build a background dossier

Everything known about a subject, in one place.

Build a Deep Research dossier on [person, company, or topic]: the key facts, the timeline, the notable details, and the open questions, drawn from current public sources and cited.

Long documents and data

Gemini handles very large inputs, up to around a million tokens of context, so it can take in long reports, contracts, and datasets in one go.

11. Summarize a long document

The gist of something you don't have time to read.

Here's a long document: [paste or attach]. Summarise it into the main argument, the key points, the decisions or numbers that matter, and three questions it leaves open. Keep it under 300 words.

12. Compare two documents

What changed, and which is stronger.

Compare these two documents and tell me where they agree, where they differ, and which is stronger on [criterion]. Put the key differences in a table. Documents: [attach both].

13. Extract structured data

Pull the facts out into a table you can check.

From this report, pull every [figure, date, name, or commitment] into a clean table, with the section each one came from so I can verify it. Report: [paste or attach].

14. Analyze a spreadsheet

Patterns and oddities in your numbers.

Here's a dataset: [attach]. Tell me the three most important patterns, anything that looks off, and one chart worth making. Explain what each finding means in plain language.

15. Synthesize several sources

One brief from a stack of documents.

I'm attaching several documents on [topic]. Synthesise them into one brief: the shared conclusions, the disagreements, and what I should take away. Note which document each point came from.

Multimodal: images, audio, and screenshots

Gemini reads images, audio, and screenshots alongside text, so you can hand it things you'd normally have to describe.

16. Analyze a photo

Ask a real question about what's in front of you.

Here's a photo: [attach]. Describe what's in it, then [answer my specific question, for example identify the plant, read the label, or explain the chart]. If you're unsure, say so.

17. Rebuild a layout from a screenshot

Turn a picture of a UI back into code.

Here's a screenshot of a [UI, layout, or form]: [attach]. Describe its structure, then write the HTML and CSS to recreate it as closely as you can. Note anything you had to guess.

18. Summarize an audio recording

Notes and action items from a recording.

Here's an audio recording: [attach]. Transcribe it, then give me a summary with the key points, any decisions, and action items with who owns each one.

19. Pull text out of an image

Get the words off a sign, receipt, or page.

Here's an image with text (a sign, a receipt, or a page): [attach]. Pull out all the text accurately, keep the structure, and flag anything you can't read clearly.

20. Explain a chart or diagram

What it shows, and what it leaves out.

Here's a chart or diagram: [attach]. Explain what it shows in plain language, the main takeaway, and one thing it might be hiding or not showing.

Create and edit images

Gemini generates and edits images with its Nano Banana model. Specific prompts beat vague ones by a mile. The trick is to describe a scene the way a photographer or designer would, naming the subject, the setting, the light, and the mood, rather than tossing in a few keywords and hoping.

21. Generate an image

Describe it like a scene, not a keyword.

Create an image of [subject] doing [action] in [setting]. Style: [photo, illustration, or 3D]. Lighting: [describe]. Mood: [describe]. Make it [orientation], clean and uncluttered.

22. Edit a photo

Change one thing, keep the rest natural.

Here's a photo: [attach]. [Remove the background, change the sky to sunset, or place the product on a clean white background]. Keep the subject sharp and natural.

23. Make a product mockup

A clean shot with room for your text.

Create a clean product mockup of [product] for [use, such as an online store]: on a [surface or setting], soft even lighting, and plenty of empty space for text. No logos or words in the image.

24. Explore visual concepts

Three directions to react to.

Generate three different visual concepts for [thing, such as a poster or a scene about X]. Make each a distinct style, and note what direction each one takes.

25. Design a thumbnail or post visual

Bold, readable, and ready for your caption.

Create a bold, eye-catching image for a [platform] post about [topic]. Leave clear space where I'll add my own text later, and keep it readable as a small thumbnail. No text in the image.

Writing and communication

The everyday writing work, done faster.

26. Draft in a specific voice

A first draft that already has a tone.

Write a [type of piece] about [topic] for [audience]. Tone: [describe]. Length: [limit]. Open with a hook, make one clear point, and avoid clichés and filler.

27. Rewrite for tone

Same meaning, different feel.

Rewrite this so it sounds [clearer, warmer, more formal, or more concise], without losing the meaning or the facts. Show me the rewrite, then list what you changed and why. Text: [paste].

28. Summarize anything

The short version, point first.

Summarise this into [a single tweet, three bullets, or a one-paragraph TL;DR] for [audience], with the most important point first. Text: [paste].

29. Reply to a tricky message

Two ways to say a hard thing well.

Help me reply to this message: [paste]. I want to [say no politely, push back, or ask for more time] while keeping the relationship good. Give me two options, one short and one fuller.

30. Turn notes into a draft

Rough thoughts into something readable.

Here are my rough notes: [paste]. Turn them into a clear, structured [email, post, or doc] in my voice, keeping every point but cutting the repetition.

Work and productivity in Google Workspace

When you use Gemini inside Google Workspace, it can work with your Gmail, Drive, Docs, and Calendar. This is where its place inside Google pays off: it isn't pasting your data into a separate tool, it's already where your email and files live. These prompts assume you're signed in with access to that data.

Two techniques make this section work far better, and most prompt lists skip both. First, you can point Gemini at a specific app by typing @ in the prompt box: @Gmail, @Google Drive, @YouTube. Typing @Gmail find the invoice from [supplier] last month forces Gemini to actually read your inbox rather than guess, and @YouTube summarise [video link] pulls the transcript so you get the content of a video without watching it. Second, Personal Context (the setting called "Instructions for Gemini") is Gemini's version of a saved system prompt: set your role, preferred tone, and format once, and it applies to every future chat. Between them they save the re-explaining that makes people give up on the Workspace features.

31. Triage your inbox

A clear read on what actually needs you.

Look at my recent emails about [topic or sender] and summarise where things stand, what needs a reply, and what I can ignore. List the replies in priority order.

32. Draft a reply from context

A response that remembers the thread.

Draft a reply to the latest email from [person], using what we've already discussed. Match a [professional but friendly] tone, cover [the points], and keep it short.

33. Prep for a meeting

Pull the pieces together before you walk in.

I have a meeting about [topic] today. Pull together what I need: the relevant emails, any related doc in my Drive, and a short agenda with the decisions we need to make.

34. Build a doc from notes

A clean document, structured for you.

Turn these notes into a clean document: [paste]. Use clear headings, a short summary at the top, and a checklist of next steps at the bottom.

35. Work in a spreadsheet

A formula, explained so you can adjust it.

In a sheet with columns [list them], help me [write a formula to do X, summarise the data, or find duplicates]. Explain the formula so I can change it myself later.

Learning, coding, and building

Understand something, write some code, or build a small tool.

36. Explain a hard concept

Plain language, with a way to check you got it.

Explain [concept] in plain language, as if to a smart beginner. Give me one analogy and one real example, then check my understanding with two questions.

37. Plan how to learn a skill

A realistic path from where you are.

I want to learn [skill] in [timeframe], starting from [level]. Build me a realistic week-by-week plan with what to practise, a few free resources, and a small project at the end.

38. Write or debug code

Working code, with an edge case to test.

Write a [language] [function or script] that [does X]. Explain how it works in comments, then suggest one edge case I should test. If I paste an error afterwards, help me fix it.

39. Build a custom Gem

A reusable assistant for a job you repeat.

Help me design a Gem (a reusable Gemini assistant) for [recurring task]. Write the instructions it should follow, the tone it should use, and what it should always ask me for. Keep it tight.

40. Prototype something in Canvas

Build a small working tool, then refine it.

Using Canvas, help me build a simple [tool, such as a budget calculator, a quiz, or a landing page] that does [function]. Start basic, show me the result, and we'll refine it together.

Video, music, and agentic tasks (new in 2026)

Google added generative video and music to Gemini during 2026, and made its fast free model, now Gemini 3.6 Flash, better at multi-step agentic work. These features often sit on a paid tier, so check what your account has before you count on them.

41. Generate a short video

A clip from a description, not a shoot.

Create a short video of [subject] doing [action] in [setting]. Style: [cinematic, animation, or realistic]. Mood and lighting: [describe]. Keep it a few seconds, and add fitting background sound.

42. Turn a script into a storyboard clip

Plan the shots, then watch them move.

Here's a short script or idea: [paste]. Break it into three or four shots with a one-line description each, then generate a short video that follows them. Keep the look consistent across shots.

43. Create background music or a jingle

A track built to a mood and a length.

Create a [length] instrumental track for [use, such as a video intro or a focus playlist]. Mood: [describe]. Tempo: [slow, medium, or upbeat]. No vocals, and keep it loopable.

44. Run a multi-step task in one prompt

Let the agentic model plan and finish the job.

Plan and complete this in one go: [goal, such as research X, draft a summary, then turn it into five slides]. Work through the steps yourself, show me your plan first, and flag anything you need from me.

45. Turn a document into a narrated recap

A watchable summary of something long.

Here's a long document: [paste or attach]. Pull out the five key points, write a short script that explains them in plain language, and turn it into a narrated video clip I can share.

Strong on current facts, but still check it

Live Google Search is what makes Gemini different, and it's a real advantage: a model that can look something up will beat one guessing from memory on anything recent. But it isn't a guarantee. Gemini can still summarise a source wrong, mix up two facts, or state a number that isn't quite right, so open the links it gives you and confirm anything that matters, especially figures, names, and dates. The research and Deep Research prompts are the easiest to trust precisely because they show their sources, so use that: when an answer cites something, the fastest check is to click through and read the line it's leaning on.

A few other limits worth knowing. Image generation is powerful but imperfect, and you're responsible for how you use what it makes, including anyone's likeness or brand. The Workspace prompts only work when you're signed in to an account that has the emails and documents you're asking about. And Google changes the models, features, and limits often, so what's free or available today may shift; the details here reflect July 2026.

What this post does not cover

These prompts are a starting point, not a guarantee of accuracy, and they aren't legal, financial, or medical advice. What Gemini can do depends on your plan, your region, and which account you're signed in with, so the free, AI Pro, and Ultra tiers won't all behave the same. Features and model names move quickly; confirm the current details before you rely on them. For the thinking behind prompts like these, see the prompt engineering basics explainer, browse the free prompt library for more, or if you're studying, our Gemini prompts for students focuses on essays, revision, and exams.

Sources

1. TechCrunch: Google's Gemini app has surpassed 750 million monthly active users
2. Google: Gemini Deep Research overview
3. Google: Use Deep Research in Gemini Apps (Help Center)
4. Google DeepMind, Gemini models (current model names: 3.6 Flash, 3.1 Pro, Nano Banana, Veo, and Lyria)