![Charles_SEO's tweet photo. BRAND NEW AI SEO TIP THAT COSTS $0 AND TAKES UNDER 15 MINUTES 🚨
Use MY EXACT PROMPT in any (good) AI - Then upload or paste your page, and the AI will split it into ~300 token chunks the way a retriever would, then checks each chunk in isolation: Does the claim survive without the surrounding context?
Because most AI search doesn't read your page - It's retrieving passages (chunks) from 5-50 different pages...
Your content gets split into blocks before retrieval, then the model often pulls the chunks that match the query and ignores everything else.
So if your claim sits in one chunk and the evidence supporting it sits in the next one, the retriever pulls ONE WITHOUT THE OTHER.
Half a fact, no context = Unusable and usually not cited.
And your rankings look completely fine while it's happening 🙃
Here's how to check in 15 minutes, for free:
STEP 1 - Pick a page that SHOULD be getting cited
Not your homepage - Pick a page that already ranks well, answers a real question, and isn't getting the love it should be from AI Overviews or chatbots.
That gap is usually a chunking problem, not a content problem.
STEP 2 - Grab the actual text
Copy the body content - No nav, no footer, no sidebar - just what a retriever would actually work with.
STEP 3 - Run this EXACT prompt
Paste it into Claude, ChatGPT, Grok or Gemini along with your content:
You are simulating how a retrieval system processes a web page for AI search.
Take the content below and split it into chunks the way a retriever would: roughly 250-400 tokens each, breaking at natural heading and paragraph boundaries.
Then evaluate EVERY chunk in complete isolation, as if it were the only text you had ever seen. For each one, output:
1. CHUNK NUMBER and the heading it sits under
2. STANDS ALONE: Yes / Partially / No
3. WHAT IT CLAIMS: the main assertion, in one sentence. Write "nothing assertable" if there isn't one.
4. IS THE CLAIM SUPPORTED INSIDE THIS CHUNK: Yes / No - is the evidence, number, example or reasoning present in this same chunk, or does it live elsewhere?
5. ORPHANED PRONOUNS: list every "it / they / this / the company / we" where the entity it refers to is NOT named inside this chunk
6. ENTITY PRESENT: is the main subject explicitly named in this chunk, or only implied?
7. CITABLE AS-IS: could a model quote or paraphrase this chunk in an answer without needing any other part of the page? Yes / No
After the chunk-by-chunk breakdown, give me:
- The 3 WEAKEST chunks and the specific reason each one fails
- Every case where a CLAIM and its SUPPORTING EVIDENCE ended up in different chunks
- A rewrite of the single worst chunk so that it stands completely alone
Be harsh. Do not give credit for anything a human reader would infer from the rest of the page. If the chunk does not contain it, it does not exist.
CONTENT:
[paste your page here]
STEP 4 - CITABLE AS-IS"
That column is whether you need to upgrade that chunk or not (or have a writing agent do it for you) - And in AI SEO client pages that I run this through, nearly 95% of the chunks need fixing...
STEP 5 - Fix the splits, not the writing
You're not rewriting the page from scratch, you're moving things to the correct structure, format and order:
- Pull the evidence UP into the same block as the claim
- Name the entity again at the top of each section instead of relying on "it"
- Break long sections with a heading BEFORE the claim, not after it
- Move orphaned stats out of the conclusion and next to what they prove
Same content, very similar word count, but completely different retrieval behaviour.
I ran this on my link building services page example from the other day because whilst it's ranking very well in Google, it's not doing as well in ChatGPT or Claude - And the prompt found exactly why...
Everyone in this industry is currently buying $500/mo dashboards to tell them they aren't getting cited.
None of them tell you WHY.
This does, it's FREE, and it takes 15 minutes.
Plus it can be automated too 🎩](https://pbs.twimg.com/media/HO8HGVQb0AA_Elp.png)
![Charles_SEO's tweet photo. BRAND NEW AI SEO TIP THAT COSTS $0 AND TAKES UNDER 15 MINUTES 🚨
Use MY EXACT PROMPT in any (good) AI - Then upload or paste your page, and the AI will split it into ~300 token chunks the way a retriever would, then checks each chunk in isolation: Does the claim survive without the surrounding context?
Because most AI search doesn't read your page - It's retrieving passages (chunks) from 5-50 different pages...
Your content gets split into blocks before retrieval, then the model often pulls the chunks that match the query and ignores everything else.
So if your claim sits in one chunk and the evidence supporting it sits in the next one, the retriever pulls ONE WITHOUT THE OTHER.
Half a fact, no context = Unusable and usually not cited.
And your rankings look completely fine while it's happening 🙃
Here's how to check in 15 minutes, for free:
STEP 1 - Pick a page that SHOULD be getting cited
Not your homepage - Pick a page that already ranks well, answers a real question, and isn't getting the love it should be from AI Overviews or chatbots.
That gap is usually a chunking problem, not a content problem.
STEP 2 - Grab the actual text
Copy the body content - No nav, no footer, no sidebar - just what a retriever would actually work with.
STEP 3 - Run this EXACT prompt
Paste it into Claude, ChatGPT, Grok or Gemini along with your content:
You are simulating how a retrieval system processes a web page for AI search.
Take the content below and split it into chunks the way a retriever would: roughly 250-400 tokens each, breaking at natural heading and paragraph boundaries.
Then evaluate EVERY chunk in complete isolation, as if it were the only text you had ever seen. For each one, output:
1. CHUNK NUMBER and the heading it sits under
2. STANDS ALONE: Yes / Partially / No
3. WHAT IT CLAIMS: the main assertion, in one sentence. Write "nothing assertable" if there isn't one.
4. IS THE CLAIM SUPPORTED INSIDE THIS CHUNK: Yes / No - is the evidence, number, example or reasoning present in this same chunk, or does it live elsewhere?
5. ORPHANED PRONOUNS: list every "it / they / this / the company / we" where the entity it refers to is NOT named inside this chunk
6. ENTITY PRESENT: is the main subject explicitly named in this chunk, or only implied?
7. CITABLE AS-IS: could a model quote or paraphrase this chunk in an answer without needing any other part of the page? Yes / No
After the chunk-by-chunk breakdown, give me:
- The 3 WEAKEST chunks and the specific reason each one fails
- Every case where a CLAIM and its SUPPORTING EVIDENCE ended up in different chunks
- A rewrite of the single worst chunk so that it stands completely alone
Be harsh. Do not give credit for anything a human reader would infer from the rest of the page. If the chunk does not contain it, it does not exist.
CONTENT:
[paste your page here]
STEP 4 - CITABLE AS-IS"
That column is whether you need to upgrade that chunk or not (or have a writing agent do it for you) - And in AI SEO client pages that I run this through, nearly 95% of the chunks need fixing...
STEP 5 - Fix the splits, not the writing
You're not rewriting the page from scratch, you're moving things to the correct structure, format and order:
- Pull the evidence UP into the same block as the claim
- Name the entity again at the top of each section instead of relying on "it"
- Break long sections with a heading BEFORE the claim, not after it
- Move orphaned stats out of the conclusion and next to what they prove
Same content, very similar word count, but completely different retrieval behaviour.
I ran this on my link building services page example from the other day because whilst it's ranking very well in Google, it's not doing as well in ChatGPT or Claude - And the prompt found exactly why...
Everyone in this industry is currently buying $500/mo dashboards to tell them they aren't getting cited.
None of them tell you WHY.
This does, it's FREE, and it takes 15 minutes.
Plus it can be automated too 🎩](https://pbs.twimg.com/media/HO8HAuxa8AA3CfN.png)
![Charles_SEO's tweet photo. BRAND NEW AI SEO TIP THAT COSTS $0 AND TAKES UNDER 15 MINUTES 🚨
Use MY EXACT PROMPT in any (good) AI - Then upload or paste your page, and the AI will split it into ~300 token chunks the way a retriever would, then checks each chunk in isolation: Does the claim survive without the surrounding context?
Because most AI search doesn't read your page - It's retrieving passages (chunks) from 5-50 different pages...
Your content gets split into blocks before retrieval, then the model often pulls the chunks that match the query and ignores everything else.
So if your claim sits in one chunk and the evidence supporting it sits in the next one, the retriever pulls ONE WITHOUT THE OTHER.
Half a fact, no context = Unusable and usually not cited.
And your rankings look completely fine while it's happening 🙃
Here's how to check in 15 minutes, for free:
STEP 1 - Pick a page that SHOULD be getting cited
Not your homepage - Pick a page that already ranks well, answers a real question, and isn't getting the love it should be from AI Overviews or chatbots.
That gap is usually a chunking problem, not a content problem.
STEP 2 - Grab the actual text
Copy the body content - No nav, no footer, no sidebar - just what a retriever would actually work with.
STEP 3 - Run this EXACT prompt
Paste it into Claude, ChatGPT, Grok or Gemini along with your content:
You are simulating how a retrieval system processes a web page for AI search.
Take the content below and split it into chunks the way a retriever would: roughly 250-400 tokens each, breaking at natural heading and paragraph boundaries.
Then evaluate EVERY chunk in complete isolation, as if it were the only text you had ever seen. For each one, output:
1. CHUNK NUMBER and the heading it sits under
2. STANDS ALONE: Yes / Partially / No
3. WHAT IT CLAIMS: the main assertion, in one sentence. Write "nothing assertable" if there isn't one.
4. IS THE CLAIM SUPPORTED INSIDE THIS CHUNK: Yes / No - is the evidence, number, example or reasoning present in this same chunk, or does it live elsewhere?
5. ORPHANED PRONOUNS: list every "it / they / this / the company / we" where the entity it refers to is NOT named inside this chunk
6. ENTITY PRESENT: is the main subject explicitly named in this chunk, or only implied?
7. CITABLE AS-IS: could a model quote or paraphrase this chunk in an answer without needing any other part of the page? Yes / No
After the chunk-by-chunk breakdown, give me:
- The 3 WEAKEST chunks and the specific reason each one fails
- Every case where a CLAIM and its SUPPORTING EVIDENCE ended up in different chunks
- A rewrite of the single worst chunk so that it stands completely alone
Be harsh. Do not give credit for anything a human reader would infer from the rest of the page. If the chunk does not contain it, it does not exist.
CONTENT:
[paste your page here]
STEP 4 - CITABLE AS-IS"
That column is whether you need to upgrade that chunk or not (or have a writing agent do it for you) - And in AI SEO client pages that I run this through, nearly 95% of the chunks need fixing...
STEP 5 - Fix the splits, not the writing
You're not rewriting the page from scratch, you're moving things to the correct structure, format and order:
- Pull the evidence UP into the same block as the claim
- Name the entity again at the top of each section instead of relying on "it"
- Break long sections with a heading BEFORE the claim, not after it
- Move orphaned stats out of the conclusion and next to what they prove
Same content, very similar word count, but completely different retrieval behaviour.
I ran this on my link building services page example from the other day because whilst it's ranking very well in Google, it's not doing as well in ChatGPT or Claude - And the prompt found exactly why...
Everyone in this industry is currently buying $500/mo dashboards to tell them they aren't getting cited.
None of them tell you WHY.
This does, it's FREE, and it takes 15 minutes.
Plus it can be automated too 🎩](https://pbs.twimg.com/media/HO8GLOfasAAtSrN.png)
Olivia
Online
LiBinbin
@BinbingLi
Proficient in the Ctrl+C / Ctrl+V language ⌨️
Joined July 2025
415 Following
295 Followers
110 Posts
@goodhunt I’m really jealous.
@MemeupMoon 此时的马斯克belike
@LukeShaw_04 越南的美食旁边都会放一盘生菜吗👀
https://t.co/SBd5oRPhbv
https://t.co/SBd5oRPhbv
@ThanksuTrump 国家的大局观一直很超前
https://t.co/tDWYOusa9z
@xuy920 越南的美食看着真有食欲啊，全是我爱吃的
I found a site that solves pretty much every subtle UI animation problem.
In a crowded market, tiny micro-interactions make users fall in love with your app. This tool curates real UI magic you can steal for inspiration.
Link: https://t.co/QPjvA2bl25
@ericsocrat 一起加油兄弟
这个文章geo提升技巧真不错呀，我已经放进gem里面了
prompt：
You are simulating how a retrieval system processes a web page for AI search.
Take the content below and split it into chunks the way a retriever would: roughly 250-400 tokens each, breaking at natural heading and paragraph boundaries.
Then evaluate EVERY chunk in complete isolation, as if it were the only text you had ever seen. For each one, output:
1. CHUNK NUMBER and the heading it sits under
2. STANDS ALONE: Yes / Partially / No
3. WHAT IT CLAIMS: the main assertion, in one sentence. Write "nothing assertable" if there isn't one.
4. IS THE CLAIM SUPPORTED INSIDE THIS CHUNK: Yes / No - is the evidence, number, example or reasoning present in this same chunk, or does it live elsewhere?
5. ORPHANED PRONOUNS: list every "it / they / this / the company / we" where the entity it refers to is NOT named inside this chunk
6. ENTITY PRESENT: is the main subject explicitly named in this chunk, or only implied?
7. CITABLE AS-IS: could a model quote or paraphrase this chunk in an answer without needing any other part of the page? Yes / No
After the chunk-by-chunk breakdown, give me:
- The 3 WEAKEST chunks and the specific reason each one fails
- Every case where a CLAIM and its SUPPORTING EVIDENCE ended up in different chunks
- A rewrite of the single worst chunk so that it stands completely alone
Be harsh. Do not give credit for anything a human reader would infer from the rest of the page. If the chunk does not contain it, it does not exist.
CONTENT:
[paste your page here]
![BinbingLi's tweet photo. 这个文章geo提升技巧真不错呀，我已经放进gem里面了
prompt：
You are simulating how a retrieval system processes a web page for AI search.
Take the content below and split it into chunks the way a retriever would: roughly 250-400 tokens each, breaking at natural heading and paragraph boundaries.
Then evaluate EVERY chunk in complete isolation, as if it were the only text you had ever seen. For each one, output:
1. CHUNK NUMBER and the heading it sits under
2. STANDS ALONE: Yes / Partially / No
3. WHAT IT CLAIMS: the main assertion, in one sentence. Write "nothing assertable" if there isn't one.
4. IS THE CLAIM SUPPORTED INSIDE THIS CHUNK: Yes / No - is the evidence, number, example or reasoning present in this same chunk, or does it live elsewhere?
5. ORPHANED PRONOUNS: list every "it / they / this / the company / we" where the entity it refers to is NOT named inside this chunk
6. ENTITY PRESENT: is the main subject explicitly named in this chunk, or only implied?
7. CITABLE AS-IS: could a model quote or paraphrase this chunk in an answer without needing any other part of the page? Yes / No
After the chunk-by-chunk breakdown, give me:
- The 3 WEAKEST chunks and the specific reason each one fails
- Every case where a CLAIM and its SUPPORTING EVIDENCE ended up in different chunks
- A rewrite of the single worst chunk so that it stands completely alone
Be harsh. Do not give credit for anything a human reader would infer from the rest of the page. If the chunk does not contain it, it does not exist.
CONTENT:
[paste your page here]](https://pbs.twimg.com/media/HPAGmZvWUAAWQAQ.png)
BRAND NEW AI SEO TIP THAT COSTS $0 AND TAKES UNDER 15 MINUTES 🚨
Use MY EXACT PROMPT in any (good) AI - Then upload or paste your page, and the AI will split it into ~300 token chunks the way a retriever would, then checks each chunk in isolation: Does the claim survive without the surrounding context?
Because most AI search doesn't read your page - It's retrieving passages (chunks) from 5-50 different pages...
Your content gets split into blocks before retrieval, then the model often pulls the chunks that match the query and ignores everything else.
So if your claim sits in one chunk and the evidence supporting it sits in the next one, the retriever pulls ONE WITHOUT THE OTHER.
Half a fact, no context = Unusable and usually not cited.
And your rankings look completely fine while it's happening 🙃
Here's how to check in 15 minutes, for free:
STEP 1 - Pick a page that SHOULD be getting cited
Not your homepage - Pick a page that already ranks well, answers a real question, and isn't getting the love it should be from AI Overviews or chatbots.
That gap is usually a chunking problem, not a content problem.
STEP 2 - Grab the actual text
Copy the body content - No nav, no footer, no sidebar - just what a retriever would actually work with.
STEP 3 - Run this EXACT prompt
Paste it into Claude, ChatGPT, Grok or Gemini along with your content:
You are simulating how a retrieval system processes a web page for AI search.
Take the content below and split it into chunks the way a retriever would: roughly 250-400 tokens each, breaking at natural heading and paragraph boundaries.
Then evaluate EVERY chunk in complete isolation, as if it were the only text you had ever seen. For each one, output:
1. CHUNK NUMBER and the heading it sits under
2. STANDS ALONE: Yes / Partially / No
3. WHAT IT CLAIMS: the main assertion, in one sentence. Write "nothing assertable" if there isn't one.
4. IS THE CLAIM SUPPORTED INSIDE THIS CHUNK: Yes / No - is the evidence, number, example or reasoning present in this same chunk, or does it live elsewhere?
5. ORPHANED PRONOUNS: list every "it / they / this / the company / we" where the entity it refers to is NOT named inside this chunk
6. ENTITY PRESENT: is the main subject explicitly named in this chunk, or only implied?
7. CITABLE AS-IS: could a model quote or paraphrase this chunk in an answer without needing any other part of the page? Yes / No
After the chunk-by-chunk breakdown, give me:
- The 3 WEAKEST chunks and the specific reason each one fails
- Every case where a CLAIM and its SUPPORTING EVIDENCE ended up in different chunks
- A rewrite of the single worst chunk so that it stands completely alone
Be harsh. Do not give credit for anything a human reader would infer from the rest of the page. If the chunk does not contain it, it does not exist.
CONTENT:
[paste your page here]
STEP 4 - CITABLE AS-IS"
That column is whether you need to upgrade that chunk or not (or have a writing agent do it for you) - And in AI SEO client pages that I run this through, nearly 95% of the chunks need fixing...
STEP 5 - Fix the splits, not the writing
You're not rewriting the page from scratch, you're moving things to the correct structure, format and order:
- Pull the evidence UP into the same block as the claim
- Name the entity again at the top of each section instead of relying on "it"
- Break long sections with a heading BEFORE the claim, not after it
- Move orphaned stats out of the conclusion and next to what they prove
Same content, very similar word count, but completely different retrieval behaviour.
I ran this on my link building services page example from the other day because whilst it's ranking very well in Google, it's not doing as well in ChatGPT or Claude - And the prompt found exactly why...
Everyone in this industry is currently buying $500/mo dashboards to tell them they aren't getting cited.
None of them tell you WHY.
This does, it's FREE, and it takes 15 minutes.
Plus it can be automated too 🎩
![Charles_SEO's tweet photo. BRAND NEW AI SEO TIP THAT COSTS $0 AND TAKES UNDER 15 MINUTES 🚨
Use MY EXACT PROMPT in any (good) AI - Then upload or paste your page, and the AI will split it into ~300 token chunks the way a retriever would, then checks each chunk in isolation: Does the claim survive without the surrounding context?
Because most AI search doesn't read your page - It's retrieving passages (chunks) from 5-50 different pages...
Your content gets split into blocks before retrieval, then the model often pulls the chunks that match the query and ignores everything else.
So if your claim sits in one chunk and the evidence supporting it sits in the next one, the retriever pulls ONE WITHOUT THE OTHER.
Half a fact, no context = Unusable and usually not cited.
And your rankings look completely fine while it's happening 🙃
Here's how to check in 15 minutes, for free:
STEP 1 - Pick a page that SHOULD be getting cited
Not your homepage - Pick a page that already ranks well, answers a real question, and isn't getting the love it should be from AI Overviews or chatbots.
That gap is usually a chunking problem, not a content problem.
STEP 2 - Grab the actual text
Copy the body content - No nav, no footer, no sidebar - just what a retriever would actually work with.
STEP 3 - Run this EXACT prompt
Paste it into Claude, ChatGPT, Grok or Gemini along with your content:
You are simulating how a retrieval system processes a web page for AI search.
Take the content below and split it into chunks the way a retriever would: roughly 250-400 tokens each, breaking at natural heading and paragraph boundaries.
Then evaluate EVERY chunk in complete isolation, as if it were the only text you had ever seen. For each one, output:
1. CHUNK NUMBER and the heading it sits under
2. STANDS ALONE: Yes / Partially / No
3. WHAT IT CLAIMS: the main assertion, in one sentence. Write "nothing assertable" if there isn't one.
4. IS THE CLAIM SUPPORTED INSIDE THIS CHUNK: Yes / No - is the evidence, number, example or reasoning present in this same chunk, or does it live elsewhere?
5. ORPHANED PRONOUNS: list every "it / they / this / the company / we" where the entity it refers to is NOT named inside this chunk
6. ENTITY PRESENT: is the main subject explicitly named in this chunk, or only implied?
7. CITABLE AS-IS: could a model quote or paraphrase this chunk in an answer without needing any other part of the page? Yes / No
After the chunk-by-chunk breakdown, give me:
- The 3 WEAKEST chunks and the specific reason each one fails
- Every case where a CLAIM and its SUPPORTING EVIDENCE ended up in different chunks
- A rewrite of the single worst chunk so that it stands completely alone
Be harsh. Do not give credit for anything a human reader would infer from the rest of the page. If the chunk does not contain it, it does not exist.
CONTENT:
[paste your page here]
STEP 4 - CITABLE AS-IS"
That column is whether you need to upgrade that chunk or not (or have a writing agent do it for you) - And in AI SEO client pages that I run this through, nearly 95% of the chunks need fixing...
STEP 5 - Fix the splits, not the writing
You're not rewriting the page from scratch, you're moving things to the correct structure, format and order:
- Pull the evidence UP into the same block as the claim
- Name the entity again at the top of each section instead of relying on "it"
- Break long sections with a heading BEFORE the claim, not after it
- Move orphaned stats out of the conclusion and next to what they prove
Same content, very similar word count, but completely different retrieval behaviour.
I ran this on my link building services page example from the other day because whilst it's ranking very well in Google, it's not doing as well in ChatGPT or Claude - And the prompt found exactly why...
Everyone in this industry is currently buying $500/mo dashboards to tell them they aren't getting cited.
None of them tell you WHY.
This does, it's FREE, and it takes 15 minutes.
Plus it can be automated too 🎩](https://pbs.twimg.com/media/HO8HIrKbQAAXKke.png)
@zm2123 成为巨佬指日可待
上个月，7 月 6 号，英国 SEO 老手 Charles Floate 发了一篇长文，标题是《25 万美元的教训》，讲他逆向工程谷歌 18 年积攒的 SEO 心得。文章发出来没几天，阅读量快九十万，算是把这一行多年没人碰过的池子点着了。
文章开头是他自己接的一个客户。那人手里有个做了十五年的金融网站，钱投了几十万美元，流量还是往下掉。做金融的都清楚，这是谷歌盯得最紧的一类，经验、专业、权威、信任一样不能少。网站自己折腾了十四年，把主关键词推到谷歌第三页，然后就再也上不去了。
Charles 给出的办法说起来很简单，用上以后很讽刺。帮客户花一千美元，从路透社买了一条新闻稿。
那一千美元买的是借权，让路透社替你的内容站台。路透社属于金融新闻里最权威的那一档，它的页面天生带着信任背书，谷歌排序时会给它另眼相看。这一招在行内有名字，寄生 SEO，用别人的权威给自己的内容开路，跟你自己埋头打磨网站是两条完全不同的路。
结果很荒唐。新闻稿当天就排到了谷歌第一页，两个小时上到前面。而客户自己的网站，同一词，十四年，最好的成绩就是第三页。
客户投入的那几十万美元，十四年的内容、外链、技术优化，在这条新闻稿面前全成了铺垫。多年攒下的一切，比不上一篇一千美元买来的文章。文章这一段的原话值得抄下来，谷歌发排名，凭得从来只信任信任与权威，你所做的努力与专业，不在这张清单里。
他讲出的做法大多能落地：
排名靠的是权威和信任，不靠你写得多卖力。同一词，路透社两小时，你的网站十四年，就是这个意思。
搜索意图由算法定义，不由用户定义。谷歌已经把每个词背后的意图钉死了，按算法的理解来，别按自己以为的理解写。
内容别追求独特。一篇稿子八成共识，再补两成信息增量，是谷歌和 AI 都能接受的形态。为不同而不同，大多数时候没有回报。
寄生 SEO 相当好用，那条一千美元的新闻稿就是活例子，两小时上到第一页。
AI 不会验证内容真假，它只看共识。多个来源都这么说，它就认定可靠。这也是为什么共识比独特值钱。
每次算法更新的东西只是权重这一层，谷歌的底层排序思路从来没换过，变的只是几个旋钮的松紧。
结尾那句话也值得留个心，这行最贵的教训，往往是钱花在了谷歌不计算的地方。原文在下面，他写得更细，这篇总结只起了个头，具体做法还得回去读原文。
#seo #google
上个月，7 月 6 号，英国 SEO 老手 Charles Floate 发了一篇长文，标题是《25 万美元的教训》，讲他逆向工程谷歌 18 年积攒的 SEO 心得。文章发出来没几天，阅读量快九十万，算是把这一行多年没人碰过的池子点着了。
文章开头是他自己接的一个客户。那人手里有个做了十五年的金融网站，钱投了几十万美元，流量还是往下掉。做金融的都清楚，这是谷歌盯得最紧的一类，经验、专业、权威、信任一样不能少。网站自己折腾了十四年，把主关键词推到谷歌第三页，然后就再也上不去了。
Charles 给出的办法说起来很简单，用上以后很讽刺。帮客户花一千美元，从路透社买了一条新闻稿。
那一千美元买的是借权，让路透社替你的内容站台。路透社属于金融新闻里最权威的那一档，它的页面天生带着信任背书，谷歌排序时会给它另眼相看。这一招在行内有名字，寄生 SEO，用别人的权威给自己的内容开路，跟你自己埋头打磨网站是两条完全不同的路。
结果很荒唐。新闻稿当天就排到了谷歌第一页，两个小时上到前面。而客户自己的网站，同一词，十四年，最好的成绩就是第三页。
客户投入的那几十万美元，十四年的内容、外链、技术优化，在这条新闻稿面前全成了铺垫。多年攒下的一切，比不上一篇一千美元买来的文章。文章这一段的原话值得抄下来，谷歌发排名，凭得从来只信任信任与权威，你所做的努力与专业，不在这张清单里。
他讲出的做法大多能落地：
排名靠的是权威和信任，不靠你写得多卖力。同一词，路透社两小时，你的网站十四年，就是这个意思。
搜索意图由算法定义，不由用户定义。谷歌已经把每个词背后的意图钉死了，按算法的理解来，别按自己以为的理解写。
内容别追求独特。一篇稿子八成共识，再补两成信息增量，是谷歌和 AI 都能接受的形态。为不同而不同，大多数时候没有回报。
寄生 SEO 相当好用，那条一千美元的新闻稿就是活例子，两小时上到第一页。
AI 不会验证内容真假，它只看共识。多个来源都这么说，它就认定可靠。这也是为什么共识比独特值钱。
每次算法更新的东西只是权重这一层，谷歌的底层排序思路从来没换过，变的只是几个旋钮的松紧。
结尾那句话也值得留个心，这行最贵的教训，往往是钱花在了谷歌不计算的地方。原文在下面，他写得更细，这篇总结只起了个头，具体做法还得回去读原文。
#seo #google
https://t.co/fcBVVhFeYL
@kim1301111 콜드 x 알고리즘, 만나서 반가워요
@heyyehao 期待token自由的一天
Last Seen Users on Sotwe
Most Popular Users
Elon Musk
@elonmusk
241.2M followers
Barack Obama
@barackobama
119.1M followers
Cristiano Ronaldo
@cristiano
112.8M followers
Donald J. Trump
@realdonaldtrump
111.8M followers
Narendra Modi
@narendramodi
107.1M followers
Rihanna
@rihanna
98.3M followers
NASA
@nasa
92.3M followers
Justin Bieber
@justinbieber
91.5M followers
KATY PERRY
@katyperry
89M followers
Taylor Swift
@taylorswift13
82.9M followers
Lady Gaga
@ladygaga
74.4M followers
Virat Kohli
@imvkohli
71.9M followers
Kim Kardashian
@kimkardashian
70.4M followers
YouTube
@youtube
68.8M followers
Neymar Jr
@neymarjr
64.8M followers
Bill Gates
@billgates
64.6M followers
The Ellen Show
@theellenshow
62.4M followers
Selena Gomez
@selenagomez
62.1M followers
CNN
@cnn
61.8M followers
X
@x
60.8M followers