Model hub

19 Gemini prompts (copy-ready and tested)

Gemini responds well to an explicit output schema stated up front, such as a fixed set of headings, a JSON structure or a table with named columns, rather than an open-ended request for "a summary" or "a plan". Giving one or two worked examples early in the prompt — showing the input and the exact output format you expect — tends to anchor the response more reliably than a purely descriptive instruction. Gemini also handles long source documents reasonably well when you separate the material to analyse from the instructions about what to do with it, using clear headings or delimiters between the two. Where it can drift is in strictly following a numeric constraint, such as an exact word count or a fixed number of bullet points, so state the constraint clearly and check the result rather than assuming it was followed precisely. Every prompt on this page follows this structure and can be pasted directly into Gemini, or checked first in the in-browser optimiser to catch a vague instruction before you send it.

Tips for Gemini

  • State the exact output schema — headings, a table or a JSON structure — near the start of the prompt rather than describing it loosely.
  • Give one worked example of input and expected output early on, especially for tasks with a specific format.
  • Separate source material from instructions using clear headings or delimiters, particularly for long documents.
  • Check numeric constraints like word counts or bullet counts after generation, since Gemini can drift from an exact figure.

Students

  • A prompt for a genuinely clear explanation of a hard topic, built with a real analogy checked for accuracy rather than a comforting oversimplification.

    Show prompt
    The prompt
    You are a teacher who is honest about where an analogy stops being accurate.
    
    Objective: Explain {{TOPIC}} to someone with {{BACKGROUND — e.g. no technical background, a high-school level of maths}}.
    
    Context: They're trying to understand this because {{WHY THEY NEED TO KNOW THIS — e.g. deciding whether to invest, making sense of a diagnosis}}.
    
    Constraints:
    - Use one concrete analogy, and explicitly say where the analogy breaks down
    - Do not use jargon without immediately explaining it in plain terms
    - Include one worked example, not just an abstract description
    - Keep the whole explanation under 300 words
    
    Output format: Return the explanation as plain prose, with the analogy's limitation as a final short paragraph.
    
    Before finishing, check that the analogy used doesn't imply something false about the topic that could mislead the reader.
    
    View full page →
  • A prompt that produces a summary you can actually act on, with the key decisions and numbers preserved instead of smoothed away.

    Show prompt
    The prompt
    You are an analyst who summarises for people who will act on your summary, not just read it.
    
    Objective: Summarise the supplied document for {{AUDIENCE — e.g. a manager who has not read the original}}.
    
    Context: The supplied document is {{DESCRIBE THE DOCUMENT — e.g. a 20-page vendor contract}}.
    
    Constraints:
    - Keep every number, date and named decision from the original exactly as stated
    - Maximum 300 words for the summary itself
    - Flag anything ambiguous or contradictory in the source rather than resolving it yourself
    - Do not add analysis or opinion that isn't in the source document
    
    Output format: Return a short summary paragraph, then a bulleted list of key facts and figures, then a separate "Open questions" section for anything unclear.
    
    Before finishing, check that every number in your summary matches a number in the source document.
    
    Document:
    {{PASTE DOCUMENT TEXT HERE}}
    
    View full page →
  • A prompt for comparing multiple sources that surfaces where they actually agree, disagree, and simply don't cover the same ground.

    Show prompt
    The prompt
    You are a research assistant whose job is to show where sources agree and disagree, not blend them into one voice.
    
    Objective: Compare the supplied sources on the question: {{RESEARCH QUESTION}}.
    
    Context: The sources are {{BRIEFLY DESCRIBE THE SOURCES — e.g. three vendor comparison articles, two academic papers}}.
    
    Constraints:
    - State which specific source each claim comes from
    - Distinguish claims the sources agree on from ones where they conflict
    - Note anything the question asks about that none of the sources actually cover
    - Do not resolve a conflict between sources yourself — present it as unresolved
    
    Example: if Source A states a 20% cost reduction and Source B doesn't mention cost at all, that's a coverage gap, not a disagreement — don't conflate the two.
    
    Output format: Return three sections — Where Sources Agree, Where Sources Disagree, and Not Covered by Any Source — each as a Markdown list with the source cited per point.
    
    Before finishing, check that every claim is attributed to a specific source and that no source is silently left out.
    
    Sources:
    {{PASTE SOURCE TEXTS OR SUMMARIES HERE}}
    
    View full page →

Marketers

  • A prompt for a value proposition statement built on a real differentiator, with three supporting proof points instead of vague claims.

    Show prompt
    The prompt
    You are a positioning strategist writing a value proposition for {{PRODUCT OR SERVICE}}.
    
    Objective: Write one value proposition statement plus supporting proof points aimed at {{TARGET CUSTOMER}}.
    
    Context: The main alternative customers currently use is {{ALTERNATIVE — e.g. doing it manually in spreadsheets, a competitor product}}. The real, defensible difference is {{DIFFERENTIATOR — e.g. it syncs automatically, no manual export needed}}.
    
    Constraints:
    - Value proposition statement: one sentence, under 30 words, names the customer and the outcome, not just the feature
    - Exactly 3 supporting proof points, each one sentence, each tied directly to the stated differentiator
    - Do not use the words "innovative", "revolutionary", "seamless" or "cutting-edge"
    - Do not claim a ranking, award or statistic that wasn't supplied
    
    Output format: Return the statement as a single bolded line, followed by the 3 proof points as a bulleted list.
    
    Before finishing, check the statement is one sentence under 30 words, there are exactly 3 proof points, and none of the banned words appear.
    
    View full page →
  • A prompt for landing page copy structured around one offer and one action, with proof points instead of vague claims.

    Show prompt
    The prompt
    You are a conversion copywriter writing landing page copy for {{PRODUCT OR SERVICE}}.
    
    Objective: Write the hero section and three supporting sections for a landing page whose single goal is {{PRIMARY GOAL — e.g. get a free trial signup}}.
    
    Context: The visitor arrives from {{TRAFFIC SOURCE — e.g. a paid search ad for "team scheduling software"}}, so they already know roughly what the product does. The strongest proof point available is {{PROOF POINT — e.g. used by 40 independent coffee shops}}.
    
    Constraints:
    - Headline: under 12 words, states the outcome, not the feature
    - Subheadline: under 25 words
    - Exactly 3 supporting sections, each with a heading under 8 words and 1–2 sentences of body copy
    - One call-to-action button label, repeated identically at the top and bottom
    - Do not state a customer count, rating or statistic other than the proof point supplied
    
    Output format: Return Markdown with a "# Headline", a subheadline line, the CTA button label on its own line, then three "##" supporting sections, then the CTA button label again on its own line.
    
    Before finishing, check there are exactly 3 supporting sections, the CTA label is identical in both places it appears, and no statistic beyond the supplied proof point is used.
    
    View full page →

SEO

  • A prompt for rewriting an underperforming page to match what searchers actually want, without inventing new claims or facts.

    Show prompt
    The prompt
    You are an SEO editor rewriting an existing page to better match search intent.
    
    Objective: Rewrite the opening section of the page below so it matches what a searcher looking for {{TARGET KEYWORD}} actually wants to find first.
    
    Context: The likely search intent is {{INTENT — e.g. wants a direct answer before reading further}}. The current opening section is:
    {{PASTE CURRENT OPENING SECTION HERE}}
    
    Constraints:
    - Use only facts present in the current opening section — do not add statistics, claims or examples not already there
    - New opening: 80–120 words
    - Answer or address the core intent within the first 2 sentences
    - Do not change the page's stated facts, only their order and framing
    
    Output format: Return the rewritten opening as a single block of prose, followed by a one-sentence note on what was reordered and why.
    
    Before finishing, check the rewrite is 80–120 words, addresses the intent within the first 2 sentences, and introduces no fact not present in the original.
    
    View full page →
  • A structured, copy-ready prompt for getting well-organised blog posts out of ChatGPT, Claude or Gemini — with audience, tone and format built in.

    Show prompt
    The prompt
    You are an experienced content writer. Write a blog post about {{TOPIC}}.
    
    Audience: {{AUDIENCE — e.g. small business owners in the UK}}
    Goal: {{GOAL — e.g. help them decide whether X is worth it}}
    Tone: clear, practical British English. No hype, no filler phrases.
    
    Structure:
    1. A hook that names the reader's actual problem in the first two sentences
    2. 3–5 H2 sections, each answering one specific question
    3. A short, honest conclusion with one clear next step
    
    Constraints:
    - 900–1,200 words
    - Short paragraphs (max 3 sentences)
    - Include one concrete example or mini case study per section
    - If you are unsure of a fact, say so rather than inventing it
    
    Before finishing, check the draft against every constraint above and confirm the word count.
    
    View full page →
  • A prompt for a writer-ready content brief with a target word count and heading structure, grounded in the given keyword and intent.

    Show prompt
    The prompt
    You are an SEO content strategist writing a brief for a freelance writer.
    
    Objective: Write a content brief for a page targeting {{TARGET KEYWORD}}, to be handed directly to a writer.
    
    Context: Search intent is {{INTENT — e.g. comparing options before buying, learning a concept}}. The page should be about {{TARGET WORD COUNT}} words. Any notes on competitor pages already ranking: {{COMPETITOR NOTES — e.g. top results are all listicles with pricing tables}}.
    
    Constraints:
    - Exactly one H1 suggestion and 4–6 H2 suggestions, each under 10 words
    - State the target word count once, matching the figure supplied
    - Include a 1-sentence brief for what each H2 section should cover
    - Do not state a search volume, keyword difficulty or competitor domain authority; none of that data was supplied
    
    Output format: Return Markdown with "Target keyword", "Search intent" and "Target word count" as single lines, then the H1, then the H2 list with its one-sentence briefs.
    
    Before finishing, check there are 4–6 H2 suggestions, the stated word count matches what was supplied, and no invented SEO metrics appear.
    
    View full page →
  • Get title tags and meta descriptions that fit the character limits and actually reflect search intent, not just stuffed keywords.

    Show prompt
    The prompt
    You are an SEO specialist who writes for the searcher first and the algorithm second.
    
    Objective: Write a title tag and meta description for a page about {{PAGE TOPIC}} targeting the search term {{TARGET KEYWORD}}.
    
    Context: The searcher's likely intent is {{INTENT — e.g. comparing options before buying, looking for a how-to}}. The page's actual content covers: {{BRIEF CONTENT SUMMARY}}.
    
    Constraints:
    - Title tag: 50–60 characters, includes the target keyword naturally
    - Meta description: 140–155 characters, states a specific reason to click, not a generic invitation
    - Do not promise anything the page content doesn't actually cover
    - Avoid keyword stuffing — the keyword should appear once, naturally
    
    Output format: Return the title tag and meta description as two labelled lines, followed by the character count for each.
    
    Before finishing, check the character counts against the constraints above and confirm the description doesn't overpromise relative to the content summary.
    
    View full page →

Social media creators

  • A prompt for turning one blog post into platform-native social posts, built only from what the post actually says.

    Show prompt
    The prompt
    You are a social media editor repurposing existing content, not writing new material from scratch.
    
    Objective: Create {{NUMBER OF POSTS — e.g. 3}} social posts for {{PLATFORM — e.g. LinkedIn, X, Instagram}} from the blog post below.
    
    Context: The blog post is:
    {{PASTE BLOG POST TEXT OR SUMMARY HERE}}
    Example: a post drawn from a section about avoiding soap would focus only on that point, not restate the whole article.
    
    Constraints:
    - Use only points, facts and examples present in the blog post — do not add a new statistic, example or claim
    - Each post must focus on one distinct point from the post — no two posts repeat the same point
    - Match {{PLATFORM}}'s typical length and formatting conventions
    - Each post must work as a standalone piece, understandable without having read the original blog post
    
    Output format: Return each post as a numbered entry, with a one-line note on which part of the blog post it's drawn from.
    
    Before finishing, check the post count matches the number requested, no two posts repeat the same point, and every claim traces back to the supplied blog post.
    
    View full page →

Developers

  • A prompt that gets a correct, readable SQL query on the first try by giving the model your schema and edge cases up front.

    Show prompt
    The prompt
    You are a database engineer who writes SQL that's easy for the next person to read.
    
    Objective: Write a {{DIALECT — e.g. PostgreSQL, MySQL, SQLite}} query that {{DESCRIBE WHAT THE QUERY SHOULD RETURN}}.
    
    Context: {{TABLE SCHEMAS — paste the columns and types}}
    
    Constraints:
    - Use explicit column names, never SELECT *
    - Handle NULLs explicitly if the result could contain them
    - Add a one-line comment above any non-obvious join or subquery
    - Do not invent column or table names that weren't given
    
    Example: "customers who placed more than 3 orders in the last 90 days" is specific enough to write directly; "our best customers" is not.
    
    Output format: Return the query in a single fenced SQL code block, followed by one sentence explaining what it does in plain English.
    
    Before answering, check the query against the schema given and confirm every referenced column actually exists there.
    
    View full page →

Image creators

  • Turns one or two photos of a person into a labelled multi-angle reference sheet for keeping a face consistent in later AI images. Works with Gemini or ChatGPT when you attach the photo; use only photos you have permission to use.

    Show prompt
    The prompt
    Using the attached reference photo(s), create ONE image: an identity reference sheet of this exact person, laid out as a labelled {{GRID — e.g. 3x3}} grid, {{ASPECT RATIO — e.g. 16:9}}.
    
    Panels (each labelled; "left" and "right" mean the subject's own left and right):
    1. Front
    2. Three-quarter left
    3. Left profile
    4. Rear three-quarter left
    5. Straight back
    6. Rear three-quarter right
    7. Right profile
    8. Three-quarter right
    9. Close-up front portrait
    
    Identity: reproduce the person exactly as photographed, including face shape, hairline, brows, eyes, nose, lips, cheeks, jaw, chin, ears, skin tone and texture, facial hair and hairstyle. Keep natural asymmetry and apparent age unchanged. Infer unseen angles conservatively from the visible structure.
    
    Constraints:
    - Every panel must show the same person, age, hairstyle and {{CLOTHING — e.g. plain grey crew-neck t-shirt}}, framed as {{FRAMING — e.g. head-and-shoulders}} at the same camera distance.
    - Neutral expression, relaxed posture, soft even studio lighting and a plain {{BACKGROUND — e.g. light grey}} background in every panel.
    - Do not beautify, idealise or symmetrise the face, and do not change the hairstyle or facial hair.
    - Realistic photograph with natural skin texture and sharp facial detail; no illustration, anime or CGI look.
    
    Before finalising, check that the face is identical in all panels. Return only the image, with no accompanying text.
    
    View full page →
  • Generates one flat, vector-style logo concept for a brand, with the mark, colours and negative space specified instead of left to chance.

    Show prompt
    The prompt
    Generate ONE logo concept image with the following exact specification.
    
    Brand: {{BRAND NAME — e.g. Northbound Coffee Roasters}}, in the {{INDUSTRY — e.g. speciality coffee}} industry.
    
    Concept: {{MARK CONCEPT — e.g. a minimalist mountain peak merged with a coffee bean silhouette}}.
    
    Style: flat, vector-style logo mark, {{STYLE DETAIL — e.g. single-weight line work, geometric and modern}}, suitable for use at both business-card and shopfront-sign size.
    
    Colours: {{COLOUR PALETTE — e.g. deep forest green and warm cream, two colours only}}.
    
    Constraints:
    - Exactly one mark, with the brand name set beneath it in a simple sans-serif typeface, not integrated into the mark itself
    - Flat vector style only: no gradients, no drop shadows, no photographic texture, no 3D rendering
    - Background: plain white, so the mark and its negative space are both clearly visible
    - No tagline, no additional decorative elements beyond the mark and the brand name
    
    Example: a two-colour geometric mark reads clearly even shrunk to the size of a favicon; a detailed illustrative one does not.
    
    Return only the generated logo image, with no accompanying text or caption.
    
    Before generating, check the concept, style and colour count above don't conflict (e.g. a "two colours only" instruction alongside a request for a gradient).
    
    View full page →
  • Generates one portrait image with a named lighting setup and mood specified precisely, instead of leaving light direction and quality to chance.

    Show prompt
    The prompt
    Generate ONE portrait image with the following exact specification.
    
    Subject: {{SUBJECT DESCRIPTION — e.g. a middle-aged man with short grey hair, wearing a navy jumper}}.
    
    Framing: {{FRAMING — e.g. head-and-shoulders, three-quarter angle}}.
    
    Lighting setup: {{LIGHTING SETUP — e.g. Rembrandt lighting, single key light from the upper left, small triangular highlight on the shadowed cheek}}.
    
    Mood: {{MOOD — e.g. calm and contemplative}}.
    
    Background: {{BACKGROUND — e.g. dark, softly blurred neutral background}}.
    
    Constraints:
    - Follow the named lighting setup precisely, including the direction, hardness and any characteristic effect it produces (e.g. the triangular highlight in Rembrandt lighting)
    - Exactly one light source described above; do not add a second, unstated light or fill light that would change the shadow pattern
    - Realistic photographic portrait, natural skin texture, no illustration, anime or CGI look
    - No text, watermark or additional people in the frame
    
    Example: "Rembrandt lighting" specifically implies a 45-degree key light producing a small triangle of light on the shadowed cheek — a flat, evenly lit face would not count as this setup.
    
    Return only the generated portrait image, with no accompanying text or caption.
    
    Before generating, check that the lighting setup described actually produces the named effect, and that no second, unstated light source has been introduced.
    
    View full page →
  • Generates one scroll-stopping social ad image with a single focal offer and space for a headline, sized correctly for a square feed placement.

    Show prompt
    The prompt
    Generate ONE social media ad visual with the following exact specification.
    
    Product or offer: {{PRODUCT OR OFFER — e.g. a limited-time 20% discount on a skincare set}}.
    
    Subject of the image: {{IMAGE SUBJECT — e.g. the skincare set's three bottles arranged on a marble surface}}.
    
    Style: {{STYLE — e.g. bright, clean, editorial product photography style}}.
    
    Composition: subject positioned {{SUBJECT POSITION — e.g. lower two-thirds of the frame}}, with the {{CLEAR AREA — e.g. upper third}} left visually clear for headline text to be added afterwards.
    
    Constraints:
    - Aspect ratio: 1:1 (square), suitable for a feed placement
    - Exactly one product or offer shown; no competing secondary products in the frame
    - Do not render any text, prices, discount percentages or logos anywhere in the image; these will be added separately afterwards
    - Bright, evenly lit, high-contrast enough to stop a scrolling thumb, but not so busy that the product itself is hard to identify at a glance
    
    Example: a single hero product shot with clear space above it for a bold headline works; a cluttered lifestyle scene with several products does not.
    
    Return only the generated image, with no accompanying text or caption.
    
    Before generating, check that the subject position and the clear area don't overlap, and that exactly one product or offer is depicted.
    
    View full page →
  • Generates one high-contrast video thumbnail with a clear focal point and space reserved for text, sized to read at small preview size.

    Show prompt
    The prompt
    Generate ONE video thumbnail image with the following exact specification.
    
    Subject: {{SUBJECT — e.g. a person looking shocked while pointing at a laptop screen}}.
    
    Topic of the video: {{VIDEO TOPIC — e.g. a coding mistake that cost a company thousands}}.
    
    Composition: subject fills {{APPROXIMATE FRAME PERCENTAGE — e.g. 60%}} of the frame, positioned {{SUBJECT POSITION — e.g. right third of the frame}}, with the {{OPPOSITE THIRD — e.g. left third}} left visually clear for a text overlay to be added afterwards.
    
    Style: {{STYLE — e.g. bold, high-contrast, slightly exaggerated expression, saturated colours}}, aspect ratio 16:9.
    
    Constraints:
    - Exactly one clear focal point (the subject); no secondary figures or distracting background elements competing for attention
    - High contrast between the subject and the background so the image stays legible when scaled down to a small preview size
    - Do not render any text, numbers or logos anywhere in the image; any text will be added separately afterwards
    - No blur across the subject; the subject must be in sharp focus even if the background is softly blurred
    
    Example: a shocked expression with an arm pointing off-frame reads clearly even at thumbnail size; a wide, busy scene does not.
    
    Return only the generated thumbnail image, with no accompanying text or caption.
    
    Before generating, check that the subject position and the clear third don't overlap, and that the style requested wouldn't obscure the single focal point.
    
    View full page →
  • Generates one reference sheet of three sample illustrations in a single consistent style, for checking a style holds before a bigger set is made.

    Show prompt
    The prompt
    Generate ONE image: a style reference sheet containing exactly three separate illustrations, arranged side by side in a single row, all in the same illustration style.
    
    Style: {{ILLUSTRATION STYLE — e.g. flat 2D vector illustration, bold outlines, limited colour palette}}.
    
    Colour palette: {{COLOUR PALETTE — e.g. teal, coral and cream, three colours only}}.
    
    The three subjects, one per panel:
    1. {{SUBJECT 1 — e.g. a person working at a desk}}
    2. {{SUBJECT 2 — e.g. a potted plant on a windowsill}}
    3. {{SUBJECT 3 — e.g. a coffee cup on a table}}
    
    Constraints:
    - Exactly three panels, each showing only the subject named for it, with a thin dividing line between panels
    - The same line weight, colour palette, shading approach and level of detail must be used identically across all three panels
    - Plain white background in every panel, with no additional scenery beyond what's needed to depict the named subject
    - No text or labels anywhere in the image
    
    Example: if panel one uses flat colour with no shading, panels two and three must also use flat colour with no shading, not a mix of styles.
    
    Return only the generated image, with no accompanying text or caption.
    
    Before generating, check that the style, palette and detail level described would look identical, not merely similar, across all three panels.
    
    View full page →
  • Generates one photorealistic interior concept render for a specific room, with layout, materials and lighting specified before generation.

    Show prompt
    The prompt
    Generate ONE interior design concept image with the following exact specification.
    
    Room: {{ROOM TYPE — e.g. a small living room, approximately 4m x 5m}}.
    
    Design style: {{DESIGN STYLE — e.g. warm Scandinavian minimalism}}.
    
    Key furniture and layout: {{LAYOUT — e.g. a low sofa against the left wall, a round wooden coffee table centred, a floor lamp in the far right corner}}.
    
    Materials and palette: {{MATERIALS AND PALETTE — e.g. light oak flooring, off-white walls, muted sage textiles}}.
    
    Lighting: {{LIGHTING — e.g. soft natural daylight from a large window on the right, warm supplementary lamp light}}.
    
    Constraints:
    - Photorealistic render, eye-level camera angle, as if photographed for a real estate or interior design listing
    - Keep every item of furniture and its position consistent with the layout given; do not add furniture that wasn't specified
    - No people, pets or visible brand names/logos anywhere in the image
    - Walls, floor and visible structural elements must remain consistent with the room type and materials given, with no conflicting materials introduced
    
    Example: "a low sofa against the left wall" is specific enough to place directly; "some nice furniture" is not.
    
    Return only the generated image, with no accompanying text or caption.
    
    Before generating, check the layout, materials and lighting above don't conflict (e.g. a window position that would put light on the wrong wall for the lighting described).
    
    View full page →
  • A structured prompt for AI product photography that specifies lighting, composition and background instead of leaving them to chance.

    Show prompt
    The prompt
    Generate a product photo with the following exact specification.
    
    Subject: {{PRODUCT — be specific about colour, material and shape}}
    
    Composition: {{ANGLE — e.g. three-quarter view, flat lay, close-up on texture}}, product filling {{APPROXIMATE FRAME PERCENTAGE — e.g. 60%}} of the frame.
    
    Style: {{STYLE — e.g. clean studio product shot, lifestyle context, minimalist}}
    
    Lighting: {{LIGHTING — e.g. soft diffused light from upper left, single hard shadow, golden hour}}
    
    Constraints:
    - Background: {{BACKGROUND — e.g. seamless white, blurred neutral surface}}
    - No text, logos or watermarks anywhere in the image unless explicitly part of the product
    - No visible people or hands unless specified above
    - Aspect ratio: {{ASPECT RATIO — e.g. 1:1, 4:5, 16:9}}
    
    Example: for online marketplace listings, a common brief is a 1:1 seamless white background shot with soft, even lighting and no shadow.
    
    Return only the generated product image, with no accompanying text or caption.
    
    Before generating, check that the composition, lighting and background above don't conflict with each other (e.g. a hard shadow requested alongside "soft diffused light").
    
    View full page →