用Coding工具完成了一次酣畅淋漓的学习,省流可以看最下面
1️⃣起因:帮客户人工跑页面优化嫌太累,准备让AI生成一套系统提示词批量跑,开着Notion(最近在做企业知识库)就用了Notion AI(新Logo好看,但能力是真的蠢),看到结构化数据里有一个HowTo类型没见过,见图一
(带“How to xxx”的关键词基本是SEO/GEO里流量和转化率最高的长尾之一,普遍出现在Featured Snippets,PAA和AI Overview里面)
2️⃣用Lovable把提示词跑了一遍,生成一个HowTo Schema,结果没法通过Schema Markup Validator(Lovable的建议给得不错,生成复杂一点的代码还是不行),见图二
3️⃣把生成的代码和Schema.org关于HowTo Schema的Docs(
schema.org)发给Cursor看,一下就找到问题生成了正确的代码,如下:
—————————
<script type="application/ld+json">
{
"@context": "
schema.org",
"@type": "HowTo",
"name": "How to Find Influencers with Lessie AI",
"description": "Find and collaborate with influencers on YouTube, TikTok, Instagram and 10+ platforms using AI-powered search and automated outreach.",
"step": [
{
"@type": "HowToStep",
"position": 1,
"name": "Describe Your Search",
"itemListElement": {
"@type": "HowToDirection",
"text": "Enter your requirements in natural language. Example: Find tech YouTubers reviewing SaaS products with 100K+ subscribers",
"url": "
lessie.ai"
}
},
{
"@type": "HowToStep",
"position": 2,
"name": "AI Multi-Platform Search",
"itemListElement": {
"@type": "HowToDirection",
"text": "Lessie AI searches YouTube, TikTok, Instagram, LinkedIn, Substack, Spotify and analyzes follower count and engagement rate.",
"url": "
lessie.ai"
}
},
{
"@type": "HowToStep",
"position": 3,
"name": "Get Verified Contacts",
"itemListElement": {
"@type": "HowToDirection",
"text": "Review results with verified email/phone contacts (95% accuracy). Export to CRM or contact directly.",
"url": "
lessie.ai"
}
},
{
"@type": "HowToStep",
"position": 4,
"name": "Automated Outreach",
"itemListElement": {
"@type": "HowToDirection",
"text": "AI agent sends personalized emails, tracks replies, and negotiates prices based on your budget automatically.",
"url": "
lessie.ai"
}
}
]
}
</script>
—————————
PS:Lessie AI(
lessie.ai)是我客户,就是前两天说起的Agentic Search Engine to Find Anyone Instantly,一款找人Agent,想找红人,求职者/应聘者,找着玩都可以试试
4️⃣还没结束,Schema Markup Validator(
validator.schema.org)是通过了,Google Rich Results Test(
search.google.com又通不过了)(习惯是用两个以上工具检测一遍,免得返工),又把Google测试的结果给Cursor提问,他也开始胡说八道了
5️⃣又去找Google Docs没找到这个Item,想起来我做了三年SEO从来没见过这个结构化数据,试探性地提问是不是因为谷歌不再使用(毕竟代码没问题就是谷歌的问题了)
6️⃣不愧是我,确实如此,顺手让Cursor给出了官方博客的信息来源(
developers.google.com),23年底/24年初就不再使用了
7️⃣然后就要纠结要不要让客户加了,有点鸡肋,因为不是必要加的字符串特别多:totalTime,estimatedCost,tool,supply,image等快十个了
8️⃣然后又问了下Cursor,这次给的建议很中肯,但部分信息还是有问题:Bing Copilot用Schema来训练LLM(
searchengineland.com),那可以推测OpenAI会受影响;Gemini,Perplexity不确定
TL;DR
如果实在不是前端资源有限,该加的Structured Data还是得加,虽然现在GEO更多依赖于语义化文本(考验内容营销能力)而不是结构数据,但也不排除有潜在收益,毕竟做这个事成本够低
AI确实好用,换之前要把这一套代码看完起码要二十分钟;AI和AI之间亦有差距,能用贵的好的就别用便宜的;也别一味依赖,要用AI,搜索引擎和个人经验交叉验证