copywriting-cta

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

End-of-Article CTA Designer

文章结尾CTA设计工具

Designing an end-of-article CTA is a function of three inputs: the objective (what action), the audience (who reads it, in what relationship to the author), and the context (independent writing, newsletter, brand publication). Get those three right and the copy + form follow almost mechanically. Skip them and you get the universal failure mode: a generic "Subscribe for more" or "Learn More" that converts at the noise floor.

This skill runs a tight interview to capture those three inputs, then prescribes a CTA: copy (what it says), form (how it looks and sits on the page), mechanism (whether to use urgency, scarcity, curiosity, reciprocity, social proof, or none), an A/B test plan, and an accessibility check.

设计文章结尾的CTA取决于三个输入要素：目标（希望用户采取什么行动）、受众（谁会阅读，他们与作者的关系），以及场景（独立创作、新闻通讯、品牌出版物）。只要把这三个要素确定好，文案和形式几乎就水到渠成。如果忽略它们，就会陷入普遍的失败模式：使用通用的“订阅获取更多内容”或“了解更多”，转化效果低到可以忽略。

此技能会通过一系列严谨的访谈收集这三个输入要素，然后给出CTA方案：文案（内容）、形式（外观和页面布局）、机制（是否使用紧迫感、稀缺性、好奇心、互惠性、社交证明，或不使用任何机制）、A/B测试计划，以及无障碍检查。

Workflow

工作流程

Run the four steps below in order. Do not skip the interview. The user may have given partial context already; pull what's available from the conversation, then ask only for the missing pieces.

按顺序执行以下四个步骤，请勿跳过访谈环节。用户可能已经提供了部分背景信息，先从对话中提取可用信息，再仅询问缺失的部分。

Step 1 — Interview

步骤1 — 访谈

Use the

ask_user_input_v0

tool. Ask one question at a time. Do not stack questions in prose. Each question must have 2-4 tappable options. Fall back to free text only if the answer genuinely cannot be enumerated.

Ask these in order, skipping any already answered:

Q1. Article context. Options:

Personal / independent blog or essay

Newsletter / paid publication (Substack, beehiiv, Ghost, etc.)

Brand / company / content-marketing blog

Other (free text)

Q2. Primary objective. (Pick the one outcome you most want from a reader who finishes the article. If they say "all of them," push back: multiple objectives is the #1 cause of CTA failure.)

Options:

```
Newsletter / email subscription
```
```
Social follow / personal branding
```

Lead generation (download / gated asset)

```
Product or service signup / free trial
```
```
Demo or sales call booking
```
```
Direct purchase
```

Community join (Discord / Slack / forum)

Engagement (reply / comment / share / restack)

Reader support (paid subscription / tip / sponsorship)

Try-it / direct action (use the code, run the tool, fork the template, open the calculator)

```
Other (free text)
```

If the user lists more than one, ask which is primary. You can offer 1-2 secondaries later, but the primary must be singular.

Q3. Audience and relationship. Options:

First-time visitor (organic search / social)

Returning reader, not subscribed

Existing subscriber / customer

Mixed / unknown

Q4. Funnel stage. (Where is the reader mentally?) Options:

TOFU: discovery, learning, no buying intent yet

MOFU: evaluating options, comparing

BOFU: ready to act, just needs a nudge

Not applicable (no buying funnel — e.g., personal blog, journalism, hobby content)

Q5. Mechanism preference. (Only ask if a mechanism could legitimately help. See

references/mechanisms.md

. For sophisticated, skeptical, or repeat-reader audiences, default to "None / value-only" without asking.) Options:

None: value statement only

Curiosity gap ("Want to know more?")

Reciprocity (free asset first)

Discount / offer

Urgency (real deadline)

Scarcity / FOMO (limited spots)

Social proof (count / testimonial)

Capture any free-text constraints the user volunteers (length limit, brand voice, no popups, multi-language, etc.). Note them.

使用

ask_user_input_v0

工具，一次只问一个问题，不要将多个问题堆叠成段落。每个问题需提供2-4个可点击选项，只有当答案确实无法枚举时，才使用自由文本输入。

按以下顺序提问，跳过已回答的问题：

问题1. 文章场景。选项：

Personal / independent blog or essay

Newsletter / paid publication (Substack, beehiiv, Ghost, etc.)

Brand / company / content-marketing blog

Other (free text)

问题2. 核心目标。（选择你最希望完成文章的读者采取的一项行动。如果用户说“全部都要”，请拒绝：多个目标是CTA失败的首要原因。）

选项：

```
Newsletter / email subscription
```
```
Social follow / personal branding
```

Lead generation (download / gated asset)

```
Product or service signup / free trial
```
```
Demo or sales call booking
```
```
Direct purchase
```

Community join (Discord / Slack / forum)

Engagement (reply / comment / share / restack)

Reader support (paid subscription / tip / sponsorship)

Try-it / direct action (use the code, run the tool, fork the template, open the calculator)

```
Other (free text)
```

如果用户列出多个目标，请询问哪个是核心目标。之后可以提供1-2个次要目标，但核心目标必须唯一。

问题3. 受众及关系。选项：

First-time visitor (organic search / social)

Returning reader, not subscribed

Existing subscriber / customer

Mixed / unknown

问题4. 漏斗阶段。（读者的心理状态如何？）选项：

TOFU: discovery, learning, no buying intent yet

MOFU: evaluating options, comparing

BOFU: ready to act, just needs a nudge

Not applicable (no buying funnel — e.g., personal blog, journalism, hobby content)

问题5. 机制偏好。（仅当机制确实有帮助时才询问。参考

references/mechanisms.md

。对于成熟、持怀疑态度或重复阅读的受众，默认选择“None / value-only”，无需询问。）选项：

None: value statement only

Curiosity gap ("Want to know more?")

Reciprocity (free asset first)

Discount / offer

Urgency (real deadline)

Scarcity / FOMO (limited spots)

Social proof (count / testimonial)

记录用户主动提出的任何自由文本限制（如字数限制、品牌语气、禁止弹窗、多语言等）。

Step 2 — Diagnose

步骤2 — 诊断

Map the inputs to a CTA archetype. The decision logic:

context = INDEPENDENT / PERSONAL
├── objective = newsletter / email      → Archetype A: Author-signature subscribe
├── objective = try-it / direct action  → Archetype B: Inline action + source link
├── objective = reader support / tip    → Archetype C: Reader-supported funding link
├── objective = community               → Archetype D: Proof-counted community invite
├── objective = social follow           → Archetype A (variant: lead with social links)
├── objective = engagement              → Archetype E: Specific reply prompt
└── objective = product / demo          → ⚠️ FLAG. Only valid on personal/professional
                                          blog where the author IS the product
                                          (consultants, coaches, solo founders, indie devs).
                                          Frame as "if you hit this, here's how I help"
                                          — never "Book a Demo" verbatim.

context = NEWSLETTER PUBLICATION
├── objective = growth / subs           → Archetype F: Share/restack + native widget
├── objective = engagement              → Archetype E: Specific reply prompt
├── objective = paid conversion         → Archetype G: Value-gap tease
├── objective = monetization / sponsor  → Archetype H: Inline sponsor block (not bottom)
├── objective = community               → Archetype D
└── objective = direct purchase         → Archetype K (rare on newsletters; use BOFU only)

context = BRAND / CONTENT MARKETING
├── stage = TOFU                        → Archetype I: Transitional asset (lead magnet)
├── stage = MOFU                        → Archetype J: Direct + Transitional pair
├── stage = BOFU                        → Archetype K: Direct CTA + risk reversal
├── objective = community               → Archetype D
└── objective = engagement              → Archetype E (rarely the right call here)

Read

references/taxonomy.md

for the full archetype catalog with copy templates, form specs, when each works, and verbatim examples from named publications.

将输入要素映射到CTA原型，决策逻辑如下：

context = INDEPENDENT / PERSONAL
├── objective = newsletter / email      → Archetype A: Author-signature subscribe
├── objective = try-it / direct action  → Archetype B: Inline action + source link
├── objective = reader support / tip    → Archetype C: Reader-supported funding link
├── objective = community               → Archetype D: Proof-counted community invite
├── objective = social follow           → Archetype A (variant: lead with social links)
├── objective = engagement              → Archetype E: Specific reply prompt
└── objective = product / demo          → ⚠️ FLAG. Only valid on personal/professional
                                          blog where the author IS the product
                                          (consultants, coaches, solo founders, indie devs).
                                          Frame as "if you hit this, here's how I help"
                                          — never "Book a Demo" verbatim.

context = NEWSLETTER PUBLICATION
├── objective = growth / subs           → Archetype F: Share/restack + native widget
├── objective = engagement              → Archetype E: Specific reply prompt
├── objective = paid conversion         → Archetype G: Value-gap tease
├── objective = monetization / sponsor  → Archetype H: Inline sponsor block (not bottom)
├── objective = community               → Archetype D
└── objective = direct purchase         → Archetype K (rare on newsletters; use BOFU only)

context = BRAND / CONTENT MARKETING
├── stage = TOFU                        → Archetype I: Transitional asset (lead magnet)
├── stage = MOFU                        → Archetype J: Direct + Transitional pair
├── stage = BOFU                        → Archetype K: Direct CTA + risk reversal
├── objective = community               → Archetype D
└── objective = engagement              → Archetype E (rarely the right call here)

阅读

references/taxonomy.md

获取完整的原型目录，包含文案模板、形式规范、适用场景以及知名出版物的真实示例。

Step 3 — Compose the recommendation

步骤3 — 撰写推荐方案

Output the recommendation in this exact structure. Do not deviate. Do not add filler.

markdown

undefined

严格按照以下结构输出推荐方案，请勿偏离，请勿添加冗余内容。

markdown

undefined

Recommended CTA

Archetype: [letter + name from decision tree] Why this fits: [1-2 sentences naming the input combination]

Content (copy)

Headline / value line:

[exact text]

Body / proof line (1-2 lines):

[exact text]

Button copy:

[exact text]

Risk reversal / subtext (if applicable):

[exact text, or "Omit: would feel forced for this audience"]

Headline / value line:

[exact text]

Body / proof line (1-2 lines):

[exact text]

Button copy:

[exact text]

Risk reversal / subtext (if applicable):

[exact text, or "Omit: would feel forced for this audience"]

Form (structure)

Placement: [end-only / end + sticky / end + mid-article repeat]
Visual weight: [low / medium / high, with justification]
Layout: [single button / button + text link / native widget cluster / one-line signature]
Proof to co-locate: [subscriber count / star count / testimonial / named recommenders / logo wall / none]

Placement: [end-only / end + sticky / end + mid-article repeat]
Visual weight: [low / medium / high, with justification]
Layout: [single button / button + text link / native widget cluster / one-line signature]
Proof to co-locate: [subscriber count / star count / testimonial / named recommenders / logo wall / none]

Mechanism

[Named mechanism + 1 sentence on why it is appropriate, OR "None: value statement carries it. Mechanisms would erode trust for this audience."]

A/B test plan

First test: [single variable, e.g., button copy A vs. B]
Why this one first: [1 sentence]
Sample size needed: [rough estimate based on baseline traffic, or "skip A/B for now — traffic too low" with the alternative recommendation]
Next 2 tests to queue: [in priority order]

First test: [single variable, e.g., button copy A vs. B]
Why this one first: [1 sentence]
Sample size needed: [rough estimate based on baseline traffic, or "skip A/B for now — traffic too low" with the alternative recommendation]
Next 2 tests to queue: [in priority order]

Accessibility check

Color contrast: [target ratio + concrete pairing if colors known]
Touch target: [size requirement]
Semantic markup: [<button> vs. <a> vs. form]
ARIA: [only if non-obvious]
Keyboard / focus: [requirement]
Color-independence: [non-color affordance]


After printing the recommendation, list 2-3 anti-patterns the user is at risk of falling into given their inputs, directly, as a contrarian check. Pull these from `references/anti-patterns.md`.

If the user is writing in a non-English language, translate the content section into that language but keep the structure (headings, labels) in English. Honor formality cues (e.g., `tu` vs. `vous` in French, `du` vs. `Sie` in German) based on prior conversation context, and flag the choice explicitly.

Color contrast: [target ratio + concrete pairing if colors known]
Touch target: [size requirement]
Semantic markup: [<button> vs. <a> vs. form]
ARIA: [only if non-obvious]
Keyboard / focus: [requirement]
Color-independence: [non-color affordance]


输出推荐方案后，列出2-3个用户当前输入可能导致的反模式，直接点明，作为反向检查。内容取自`references/anti-patterns.md`。

如果用户使用非英语语言写作，请将内容部分翻译成该语言，但保持结构（标题、标签）为英文。根据对话上下文遵循正式程度提示（如法语中的`tu`与`vous`，德语中的`du`与`Sie`），并明确标注该选择。

Step 4 — Offer next moves

步骤4 — 提供后续建议

Suggest 2-3 follow-up directions:

Steelman the opposite. Offer to design the CTA you would recommend against — e.g., the hard-sell version on a TOFU post — so the user can see why it fails.
Variant for a different audience or platform. If the article will be cross-posted (own site + Medium + LinkedIn + a syndication network), offer to rewrite per platform.
End-to-end review. Offer to audit the rest of the article for CTA-supporting signals: author bio, related-post links, in-line proof.

建议2-3个后续方向：

反向验证：提供你不推荐的CTA设计方案（例如，在TOFU阶段的文章中使用硬推销版本），让用户了解其失败原因。
多平台适配变体：如果文章将跨平台发布（自有网站+Medium+LinkedIn+联合发布网络），可根据不同平台重写CTA。
端到端审查：提供文章其余部分的CTA支持信号审计：作者简介、相关文章链接、内联证明。

Style inheritance

风格继承

The copy templates in

references/taxonomy.md

are starting points, not finished copy. Always adapt them to:

The user's stated brand voice or any
```
<userPreferences>
```
in scope (formality, language, em-dash avoidance, length limits).
The language of the article. Output copy in the article's language; never default to English.
The publication's existing voice. If the user has prior posts visible, mirror their cadence and vocabulary.
The reader's expected level of expertise. A CTA for a beginner-finance blog uses different vocabulary than one for a quant-trading newsletter.

Never output a template verbatim if it conflicts with the user's stated style preferences.

references/taxonomy.md

中的文案模板是起点，而非最终文案。始终根据以下内容进行调整：

用户指定的品牌语气或范围内的任何
```
<userPreferences>
```
（正式程度、语言、避免使用破折号、字数限制）。
文章的语言。输出文案需与文章语言一致，切勿默认使用英文。
出版物的现有语气。如果用户有可见的往期文章，需匹配其节奏和词汇。
读者的预期专业水平。面向入门级金融博客的CTA，词汇与面向量化交易新闻通讯的CTA不同。

如果模板与用户指定的风格偏好冲突，切勿直接输出模板内容。

Reference files

参考文件

Read these as needed during diagnosis and composition. Read the relevant file in full before composing the recommendation; do not paraphrase from memory.

references/taxonomy.md
: All 11 archetypes (A through K) with copy templates, form specs, verbatim examples from named publications, and conversion expectations.
references/mechanisms.md
: When to use urgency, scarcity, FOMO, discount, curiosity, reciprocity, social proof, authority, unity. When NOT to use them.
references/ab-testing.md
: Priority order of variables to test, sample-size rules of thumb, common pitfalls, when to skip A/B testing entirely.
references/accessibility.md
: WCAG 2.2 specifics for CTA blocks: contrast ratios, touch targets, ARIA patterns, focus states, keyboard support, motion preferences.
references/anti-patterns.md
: 12 failure modes to call out by name when they apply to the user's inputs.

在诊断和撰写过程中按需阅读这些文件。撰写推荐方案前，请完整阅读相关文件，切勿凭记忆转述。

references/taxonomy.md
：全部11个原型（A至K），包含文案模板、形式规范、知名出版物的真实示例以及转化预期。
references/mechanisms.md
：何时使用紧迫感、稀缺性、FOMO、折扣、好奇心、互惠性、社交证明、权威性、统一性；何时不应使用这些机制。
references/ab-testing.md
：测试变量的优先级顺序、样本量经验法则、常见陷阱以及何时完全跳过A/B测试。
references/accessibility.md
：CTA模块的WCAG 2.2具体要求：对比度、触摸目标、ARIA模式、焦点状态、键盘支持、动效偏好。
references/anti-patterns.md
：12种失败模式，当适用于用户输入时需明确指出。

Operating principles

操作原则

One primary CTA per post. Multiple competing CTAs is the dominant failure mode (single-CTA pages convert ~30%+ better than multi-CTA pages in repeated case studies).
Match the voice of the publication. A personal-essay footer that reads like a SaaS landing page collapses credibility. A SaaS footer that reads like a casual signature converts at noise.
Specificity beats cleverness. "Get one essay a week on indie filmmaking" beats "Subscribe to our awesome newsletter." Joanna Wiebe's "I want to ___" completion test is the cleanest filter for button copy.
Proof co-located with the ask. Subscriber count, testimonial, customer logos, star count, named recommenders — whichever signal is honest for the context, place it inside or adjacent to the CTA block.
Mechanisms are tools, not garnish. Most well-written value statements need no mechanism. Add urgency, scarcity, FOMO, or discount only when the context genuinely supports them; theatrical mechanisms erode trust faster than they lift conversion.
Push back on bad asks. If the user wants "Book a Demo" at the bottom of a beginner tutorial for first-time visitors, say so. Do not produce a polished version of a CTA that will fail. Propose the alternative, explain why, then if the user still wants the original, deliver it with the failure mode flagged.

每篇文章仅设一个核心CTA：多个相互竞争的CTA是主要的失败模式（多项案例研究表明，单CTA页面的转化率比多CTA页面高约30%以上）。
匹配出版物的语气：个人随笔的页脚如果像SaaS着陆页一样，会丧失可信度；SaaS的页脚如果像随意的签名，转化率会低到可以忽略。
具体性胜过巧妙性：“每周获取一篇独立电影制作相关随笔”比“订阅我们精彩的新闻通讯”效果更好。Joanna Wiebe的“我想____”完成测试是筛选按钮文案最简洁的方法。
证明信息与请求紧邻：订阅人数、推荐语、客户标志、星级评分、知名推荐人——无论哪种信号符合场景，都应放置在CTA模块内部或相邻位置。
机制是工具，而非装饰：大多数写得好的价值声明不需要任何机制。只有当场景真正需要时，才添加紧迫感、稀缺性、FOMO或折扣；刻意的机制会比提升转化率更快地侵蚀信任。
拒绝不合理的请求：如果用户希望在面向首次访问者的入门教程底部放置“预约演示”，请明确指出问题。不要生成一个注定失败的CTA的精致版本。提出替代方案，解释原因，如果用户仍坚持原请求，则提供方案并标注失败模式。