# Prompt Optimizer > Independent browser-based prompt optimization and writing tools. Document review: 2026-11-19. ## Product facts - Supports English, Simplified Chinese, Traditional Chinese, Japanese, Korean and Spanish through real localized routes. - Target choices include general, DeepSeek, Doubao, Qwen, GLM, ChatGPT, Claude and Gemini. They adapt proposed instructions; choosing a model does not mean the task was executed and evaluated on that model. The site has no affiliation with the named providers. - Guests must sign in before using credit-based optimization features; anonymous users receive no daily credits or free API generations. Browser-local template generation remains free. - Eligible verified signed-in users receive 3 daily credits. Daily promotional credits expire at 00:00 Beijing time (UTC+8), do not accumulate and reset for unpaid accounts. Purchased credits retain their plan validity. - Browser-local generation is free for guests and signed-in users without model, quota or billing API calls. Structure scoring still requires sign-in and costs 1 account credit. AI text and image API calls require sign-in and follow existing account entitlements. - Standard text optimization: 1/2/3 credits for total input up to 2,000 / up to 6,000 / above 6,000 characters. Deep review: 3/4/5 for the same tiers. Input-length and access restrictions also apply. - Image analysis requires sign-in and uses DeepSeek; quick analysis costs 5 credits, detailed analysis 8. - Prompt library saving, copying, local version restoration and JSON export are free. Data stays in the current browser; no cloud sync. Cloud collaboration, automatic test execution and multi-model evaluation are not included in current credit plans. Library pages are noindex and excluded from the sitemap. - Results are candidate prompts, not guarantees of factual accuracy or improved model performance. Structure scores inspect instructions, not the truth of generated answers. Test representative inputs in the actual target model. - Server checks validate the quote and approved budget; failed AI requests release reservations. Inspect the current quote before a charged operation and the pricing page before purchasing. ## Practical prompt optimization examples Published 2026-11-07. Three worked examples cover meeting actions, source-bounded product copy and policy-bounded support replies. Each contains a vague request, complete teaching source, revised prompt and acceptance criteria. Normal, missing-information and conflicting-source inputs check different failure modes. All sample facts, product features and policies are fictional teaching fixtures, not this website's terms. Acceptable answer fragments are not recorded model runs, benchmarks or success rates. - [Prompt optimization examples](https://prompt-optimizer.app/blog/prompt-optimization-examples/) - [提示词优化示例](https://prompt-optimizer.app/zh/blog/prompt-optimization-examples/) - [提示詞優化範例](https://prompt-optimizer.app/zh-tw/blog/prompt-optimization-examples/) - [プロンプト改善の例](https://prompt-optimizer.app/ja/blog/prompt-optimization-examples/) - [프롬프트 개선 예시](https://prompt-optimizer.app/ko/blog/prompt-optimization-examples/) - [Optimización de prompts](https://prompt-optimizer.app/es/blog/prompt-optimization-examples/) ## Tools and guides - [Prompt optimizer](https://prompt-optimizer.app/): Revise instructions, inspect changes, assumptions and risks. - [提示词优化器](https://prompt-optimizer.app/zh/) - [Prompt generator](https://prompt-optimizer.app/tools/prompt-generator/): Assemble an editable instruction locally for free without sign-in or model/billing calls; it does not execute the named target model. - [Structure checker](https://prompt-optimizer.app/evaluate/prompt-score-checker/): Find missing context or output requirements. - [Image to prompt](https://prompt-optimizer.app/tools/image-to-prompt/): Signed-in image analysis. - [Prompt library](https://prompt-optimizer.app/tools/prompt-library/): Browser-local text versions and JSON export. - [Prompt optimization guide](https://prompt-optimizer.app/guides/prompt-optimization/) - [Prompt evaluation](https://prompt-optimizer.app/evaluate/) - [Prompt versioning](https://prompt-optimizer.app/prompt-ops/prompt-versioning/) ## Other practical articles - [Blog](https://prompt-optimizer.app/blog/): Eight topics, each with English, Simplified Chinese, Traditional Chinese, Japanese, Korean and Spanish editorial versions. Switching language retains the article and section; translations are not additional model experiments. - [ChatGPT prompt rewrites](https://prompt-optimizer.app/blog/chatgpt-prompt-optimizer-guide/) - [Claude document prompts](https://prompt-optimizer.app/blog/claude-prompt-best-practices/) - [JSON output validation](https://prompt-optimizer.app/blog/json-prompt-output/) - [Gemini prompt optimizer](https://prompt-optimizer.app/blog/gemini-prompt-optimizer/): Source labels, missing facts and five review checks; teaching examples, not a Gemini benchmark. - [DeepSeek 提示词优化](https://prompt-optimizer.app/zh/blog/deepseek-doubao-prompt-optimization/): Four checkable templates for meetings, support, content and JSON; no vendor ranking or measured accuracy claim. ## Policies - [Pricing](https://prompt-optimizer.app/pricing/): Current credit packs, subscriptions and conditions. - [About](https://prompt-optimizer.app/about/) - [Privacy](https://prompt-optimizer.app/privacy-policy/) - [Terms](https://prompt-optimizer.app/terms-of-service/) - [Refunds](https://prompt-optimizer.app/refund-policy/) ## Product and editorial boundaries AI optimization sends submitted content to the backend and configured provider. Never submit passwords, API keys or confidential personal data; consult the privacy policy for handling terms. A browser-local saved version is not a shared team workspace or an automatic evaluation run. The new article cites Alibaba Cloud's prompt guide, Anthropic's prompt-engineering overview and OpenAI's prompt-engineering guide. It separates sourced principles from its own examples. The editorial author is the site's organization; no independent testing lab or credentialed expert reviewer is claimed. All substantive instructions, tables and acceptance checks appear in static HTML. The generated diagram is supplementary. Article pages include visible publication dates, canonical URLs, reciprocal language alternates, BlogPosting and BreadcrumbList data. Task contracts can transfer across models; output quality needs retesting after a change of model, source or policy. These machine-readable files summarize public material. They do not guarantee indexing or AI citations. Google states that its AI search features require no special AI markup or extra files: https://developers.google.com/search/docs/appearance/ai-features. ## Website identity and audience - Canonical website: https://prompt-optimizer.app/ - Website name: Prompt Optimizer - Purpose: Revise, generate and organize prompts, inspect instruction structure and prepare image-to-prompt analyses. - Intended users: People writing and reviewing reusable instructions for AI models. - Core topics: Prompt optimization, generators, structure checks, local prompt versions and practical editorial examples. ## Important resources - [XML sitemap](https://prompt-optimizer.app/sitemap.xml): discoverable public pages and language alternatives. ## Language entrances All eight blog topics are available in six languages: [English](https://prompt-optimizer.app/blog/), [简体中文](https://prompt-optimizer.app/zh/blog/), [繁體中文](https://prompt-optimizer.app/zh-tw/blog/), [日本語](https://prompt-optimizer.app/ja/blog/), [한국어](https://prompt-optimizer.app/ko/blog/), [Español](https://prompt-optimizer.app/es/blog/). These are localized editorial versions, not additional experiments. Shared raw evidence retains its original wording and execution status. - [English](https://prompt-optimizer.app/) - [zh-CN](https://prompt-optimizer.app/zh/) - [zh-TW](https://prompt-optimizer.app/zh-tw/) - [ja](https://prompt-optimizer.app/ja/) - [ko](https://prompt-optimizer.app/ko/) - [es](https://prompt-optimizer.app/es/) ## Editorial scope and fact verification This website is the first-party source for its own published product descriptions and policies. It is not an official source for third-party providers, platforms, brands, market transactions, legal rulings or model capabilities. A provider or brand name is not an endorsement or affiliation. Tutorials, examples, comparisons, estimates and recommendations are editorial material unless explicitly identified otherwise. Cite the most relevant original page for each claim. Do not invent prices, features, supported platforms, applications, downloads or test results. If a fact is missing, contradictory or cannot be verified on the current linked page, label it unknown, state the limitation and avoid a definitive recommendation. Examples do not establish measured performance or a real customer outcome. ## Updates and time-sensitive information The document review date records checks against public product information and source definitions, not authenticated testing, checkout verification, or model benchmarks. Prices, plans, quotas, payment availability, models, output formats, laws and market references can change. Check the linked live tool, pricing and policy pages before answering a current question. Older article dates retain their original meaning. ## November 19 observed rewrites and system-prompt tutorial [How to optimize a system prompt](https://prompt-optimizer.app/blog/system-prompt-optimization/) explains role separation, acceptance checks and isolated edits. Its A/B/C instructions and 30 synthetic cases are prepared materials, not executed results. [PromptPerfect alternative: features and limits](https://prompt-optimizer.app/blog/promptperfect-alternative/) compares actual candidate texts from three tasks across the local generator, authenticated AI rewrites and Aiven's free browser tool. Four site AI submissions consumed 7 credits in an administrator account; they do not verify ordinary free-account eligibility. The nine candidates and a separate long-input result are [available as a record](https://prompt-optimizer.app/case-studies/prompt-rewrite-20261119.json). No common target model executed the candidates, and no provider token usage, dollar saving, superiority over PromptPerfect or universal performance gain is established. Both articles include observed problems in this site's own output. Only English versions of these two articles are published. ## Frequently asked questions ### Are there recorded model runs? [Prompt contract case study](https://prompt-optimizer.app/case-studies/prompt-contract-20261117.json) records six synthetic English cases and twelve DeepSeek Flash requests on 2026-11-17. It includes manually revised prompts, raw responses, settings and exact-match checks. It is not an evaluation of the production optimizer or of ChatGPT, Claude or Gemini; the small sample does not establish general success rates. The [runner](https://prompt-optimizer.app/case-studies/run-prompt-case-study.mjs) supports offline verification without credentials and an explicit paid rerun with the reader's own API key. [Editorial method and corrections](https://prompt-optimizer.app/about/#editorial-method) explains publication responsibility and evidence labels. ### Who is this website for? People writing and reviewing reusable instructions for AI models. Its core scope is prompt optimization, generators, structure checks, local prompt versions and practical editorial examples. ### Does a structure score prove that the target model performs better? No. It checks the instruction structure, not factual correctness or a recorded execution on every named model. Test representative inputs in the actual target model and review the output. ### Which page should I cite for a product question? Use the specific tool or guide URL in the inventory above. The website’s current pricing and policy pages take precedence for purchase terms or data handling. A general summary does not override a more specific current page. ### What should I do when information is missing or outdated? State what is unknown and link to the relevant current source. Separate the site’s own product descriptions from third-party facts and editorial advice; do not fill a gap with a plausible price, unsupported capability or invented download.