---
title: "How I'd grow Every with SEO, a full teardown | SEO Skill"
description: "An SEO teardown of Every, the AI media company behind Monologue and Cora. Where its Google traffic lands, what makes money and the fixes I'd ship."
canonical: "https://seoskill.dev/blog/how-id-grow-every"
language: "en"
---
# How I'd grow Every
Every's site has a Domain Rating of 77, level with its rivals in AI media, but most of its search traffic lands on pages that don't sell anything. This is how I'd send more of it to consulting, courses and the apps, and the new pages I'd build to win the searches its rivals get.
Written by AI and Ian NuttallX: 81.8kYouTube: 16.9kLinkedIn: 4.5k
Ian Nuttall
More than 20 years in SEO. Founder of Practical Programmatic, a programmatic SEO course with over 1,000 students, and of URL Monitor, sold for $250,000.
- [81.8kfollowers on X](https://x.com/iannuttall)
- [16.9ksubscribers on YouTube](https://www.youtube.com/@inuttall)
- [4.5kfollowers on LinkedIn](https://www.linkedin.com/in/iannuttall)
AI writes these posts. I plan each one, gather the data from real sources and review the draft before it goes live.
[See my background](https://seoskill.dev/about#ian-nuttall)
4 November 2026
[Every](https://every.to/) is an AI media company. It publishes a daily newsletter and essays, sells a subscription, runs consulting for companies adopting AI, teaches cohort courses, and builds its own apps: Monologue, Cora, Spiral and Sparkle.
So Every makes money five ways from one audience. It also has a site with a Domain Rating of 77 and about 7,800 linking sites. In this teardown I look at how to use that site to sell more consulting, courses and apps.
## Most of Every’s search traffic isn’t buying anything
Ahrefs estimates every.to gets about 7,300 visits from Google a month. This map shows where they land.
- Typos and off-topic searches3,445
- Searches for Every by name1,858
- Essays, guides and podcasts964
- Consulting and newsletter sign-up573
- Vibe Check model reviews301
- App launch posts115
Estimated Google visits to every.to by kind of page, 4 November 2026 (Ahrefs top pages, all countries). Blue pages sell something. Course pages get no measurable search traffic.
About 9% of those visits reach a page that sells something. Most of the rest come from typos and off-topic searches, or from people searching for Every by name, who would find it anyway.
Every has had between about 7,800 and 45,000 visits a month from Google over the last year, depending on the month (Ahrefs estimates). Most of the swing is in two places. The first is the homepage, which Ahrefs put at about 9,700 visits a month on 2 November and about 1,500 two days later. Every still ranks first for “every”, so Ahrefs changed its estimate, not Every’s ranking, and those visits were never buyers.
The second is a handful of pages. The clearest example is a podcast page called “Introducing How Do You Use ChatGPT?”. In September it was the biggest page on the site after the homepage, at about 10,800 visits a month. Most of them came from people typing “hatgpt”, a typo of ChatGPT.
| Label | Value (visits/month) | Note |
| --- | --- | --- |
| September 2026 | 10,845 | Biggest page after the homepage, mostly searches for "hatgpt" |
| November 2026 | 2,348 | |
Estimated monthly visits to one podcast page (Ahrefs). Typo traffic arrives quickly and leaves the same way.
It’s now down to about 2,300. Every lost very little, because someone who mistyped ChatGPT wasn’t going to hire its consultants. For each part of the site below, I check whether its search traffic reaches a page that sells something.
## AI consulting: Google’s AI answer cites Every, and the results below it don’t
Every Consulting helps companies adopt AI. It’s the part of Every where one new client is worth more than thousands of subscribers, so it’s where search traffic matters most.
These are the biggest consulting searches in the US, with monthly volumes and Every’s positions from Ahrefs.
| Search | US searches | Difficulty | Every today |
| --- | --- | --- | --- |
| ai consulting | 8,500 | 38 | #11 |
| ai consulting services | 5,800 | 28 | Not in top 100 |
| ai strategy consulting | 3,600 | 5 | Not in top 100 |
| ai consulting company | 2,300 | 46 | AI Overview, #8 |
| ai consultants | 1,500 | 31 | Not in top 20 |
| ai consulting agency | 1,400 | 51 | #11 |
| generative ai consulting | 1,200 | 18 | Not in top 100 |
| artificial intelligence consulting company | 1,100 | 50 | AI Overview, #11 |
| ai transformation consulting | 800 | 17 | #10 |
| ai consulting firm | 700 | 46 | AI Overview, #9 |
Ahrefs, US, November 2026. Every has no page for more than 10,000 other searches a month.
Every’s only consulting page has about 760 words. For “ai consulting company”, “ai consulting firm” and “artificial intelligence consulting company”, Google’s AI Overview cites it next to McKinsey, BCG and IBM. In the ordinary results under that answer, it ranks between eighth and eleventh. For “ai consulting”, the biggest search, it’s 11th, at the top of page two.
So Google already uses the page as a source for its answer. It needs more depth and more links to rank with it, and Every needs pages for the searches it has nothing for yet.
The biggest of those is “ai strategy consulting”, with 3,600 searches a month and a difficulty of 5. These are the pages Google shows for it now.
ai strategy consulting
3,600 searches a month
1. 2bcg.comConsultancyDR 91
2. 3reddit.comForum threadDR 95
3. 5quandarypeak.comConsultancyDR 49
4. 6linkedin.comLinkedIn postDR 99
5. 7thehackettgroup.comGlossaryDR 75
6. 8medium.comBlog postDR 94
7. 10cgi.comConsultancyDR 82
8. ×every.toNot ranking in these results
DR is [Domain Rating by Ahrefs](https://ahrefs.com/website-authority-checker).
Results for "ai strategy consulting" (Ahrefs, US). Below BCG, it's a forum thread, a small consultancy, a LinkedIn post, a glossary and a Medium post.
A forum thread and a Medium post on page one usually means no firm with real authority has written a good page for the search. Every has the authority, the client stories and the consultants to write one.
“ai consulting services” (5,800) is harder: EY, BCG and IBM hold the top spots, with “top 10 AI consulting companies” lists in between. AI answers quote those lists too (the AI search section below shows which ones). To stay in Google’s AI Overview, Every needs to be named in the lists that rank.
I’d turn consulting into a small hub: a page each for strategy, implementation and team training, each with a client story, linked from the main page. I’d keep the main page’s address and copy, since Google already cites it, and add the client stories and a short “how an engagement works” section that the hub pages link back to.
## Courses: compete where a small free site already ranks
Every runs cohort camps for Claude Code, Cursor, Droid and compound engineering. Each run gets its own address and announcement post. The Claude Code camp has had at least three runs, at three different addresses, and Every’s courses page sends people to a general events page.
These are the course searches with the most volume, and the sites that rank for them.
| Search | US searches | Difficulty | Who ranks | Can Every compete? |
| --- | --- | --- | --- | --- |
| ai courses | 7,700 | 51 | Coursera, Google, DataCamp, Harvard | No |
| claude code tutorial | 2,800 | 17 | Anthropic, YouTube, a Substack post | Yes |
| compound engineering | 1,600 | 12 | Every's GitHub plugin, Every's guide | Already does |
| claude code course | 1,000 | 59 | Anthropic, YouTube, Reddit | Yes, below Anthropic |
| learn claude code | 1,000 | 52 | Anthropic, Reddit, freeCodeCamp | Yes |
| ai training for employees | 500 | 16 | HR and training vendors | Yes |
Ahrefs, US. "ai courses" is a catalogue search won by sites with thousands of courses. The Claude Code searches are won by Anthropic, videos and one small independent site.
I’d skip “ai courses”. The sites that rank for it, such as Coursera and Harvard, list thousands of courses. Every has five camps, so it can’t offer the catalogue those searchers want.
Every has a much better chance with the Claude Code searches.
claude code course
1,000 searches a month
1. 1academy.claude.comAnthropicDR 92
2. 3anthropic.skilljar.comAnthropicDR 83
3. 4academy.claude.comAnthropicDR 92
4. 5reddit.comForum threadDR 95
5. 7anthropic.skilljar.comAnthropicDR 83
6. 9linkedin.comCourse catalogueDR 99
7. 10ccforeveryone.comFree courseDR 32
8. ×every.toNot ranking in these results
DR is [Domain Rating by Ahrefs](https://ahrefs.com/website-authority-checker).
Organic results for "claude code course" (Ahrefs, US). Videos also sit at position 2. The last result is a free course site with a Domain Rating of 32.
Anthropic’s own academy holds the top spots. Below it are YouTube videos, Reddit threads and ccforeveryone.com, a free course site with a Domain Rating of 32. Every’s DR is 77, it teaches the subject, and it has three cohorts of students. A permanent Claude Code course page, with reviews from past cohorts and the next start date, could rank just below Anthropic.
The same page can answer “claude code tutorial” (2,800 searches, difficulty 17) with a free first lesson. A page for “ai training for employees” would help the consulting team as well as the courses team. HR and training vendors rank for that search now.
## The model reviews are close to ranking
Every reviews each new AI model in its Vibe Check series, which is what someone searching for a new model wants to read. Their titles are written for the brand. They start with “Vibe Check:”, and the GPT-5 review’s title is just “GPT-5”.
People search for the model name. “gpt 5.6 sol” gets about 2,500 US searches a month, and Every’s review already ranks seventh for “gpt sol”. “claude vs chatgpt” gets 35,000 (Ahrefs, November 2026), and Every has no page for it. The new content section below covers those comparisons.
Today
Everyhttps://every.to › vibe-check › gpt-5-6-sol
Vibe Check: GPT-5.6 Sol Is Our Favorite Model to Collaborate With
Sol is fast, resourceful, and unusually easy to steer—but Fable still gets the assignments we want to hand off completely
Title 65 characters. Description 121 characters.
After the fix
Everyhttps://every.to › vibe-check › gpt-5-6-sol
GPT-5.6 Sol review: our favorite model to work with
Every's hands-on GPT-5.6 Sol review: fast, resourceful and easy to steer, and the jobs where we still hand work to Fable instead.
Title 51 characters. Description 129 characters.
The GPT-5.6 Sol review in Google today, and with a title and description written for someone searching for the model. The page headline keeps the Vibe Check name.
I’d keep the Vibe Check name, because readers know it. I’d change the title tag that Google shows and name the model in the first paragraph and the subheadings. Then I’d add a hub page for each model family (GPT, Claude, Gemini) that links to every review, so each new review gets a link from the hub on the day it’s published.
## Every’s links go to the wrong pages
I only found this because I crawled every.to for links to Every’s other sites.
I crawled 600 every.to pages and pulled out every link to Every’s own apps. There are a lot of them, because the newsletter template promotes the products.
1. Cora homepageAlready ranks for its own name: 449 links
2. Sparkle homepageAlready ranks for its own name: 440 links
3. Monologue homepageIts comparison pages get no links: 278 links
4. Spiral homepageAlready ranks for its own name: 178 links
5. Every other product URL combinedApp logins, changelogs and a few feature pages: 28 links
Links from 600 crawled every.to pages to Every's own apps. The other 28 go to app logins, changelogs and a few feature pages. None goes to a comparison or use-case page.
Nearly all of these 1,300 links go to four homepages that already rank for their own names, and the link text is just the product name. Monologue has comparison pages against Wispr Flow and Superwhisper that get no links at all. Four posts also link to monologue.co, a parked domain, which looks like a typo for monologue.to. The [Monologue teardown](https://seoskill.dev/blog/how-id-grow-monologue) lists them.
Most of these links come from the newsletter template, so changing the template fixes them in every post at once. I’d point each product mention at the page that sells to that reader, with link text that says what’s there (“dictation app for Mac”, “AI email assistant”).
## New content: the searches Every’s rivals win
Every has the authority to rank for much more than it does. Its Domain Rating is level with the sites it competes with for readers. Maven, Simon Willison and Lenny get two to four times its search traffic, and The Rundown gets 21 times.
1. therundown.aiDomain Rating 72. One page per AI tool: 151,648 visits
2. maven.comDomain Rating 79. Cohort courses, like Every's camps: 25,908 visits
3. simonwillison.netDomain Rating 85. A post per model and small free tools: 17,021 visits
4. lennysnewsletter.comDomain Rating 79. Its biggest page is a hands-on review of Claude Design: 15,722 visits
5. every.toDomain Rating 77: 7,267 visits
6. oneusefulthing.orgDomain Rating 78. Ranks first for "how to use ai": 7,158 visits
Estimated monthly Google visits, all countries, 4 November 2026 (Ahrefs). The Rundown AI gets most of its visits from a directory with one page per AI tool.
The US searches these sites rank for, where Every has no page in the top 100, fall into three groups, and each one leads to something Every sells.
What the winning pages have in common is that they’re specific. Lenny’s biggest page is “What Claude Design is actually good for”, which ranks fifth for “claude design” (43,000 US searches a month). Ethan Mollick’s “Using AI Right Now: A Quick Guide” ranks second for “how to use ai” and Google’s AI Overview cites it. Simon Willison writes a post for each new model, and ranks for the model’s name. Every does that kind of work in its Vibe Checks and camps, but on pages written for people who already subscribe.
### 1\. Model comparisons: about 89,000 searches, and Every already does the testing
People choosing between AI models search for the pair. Every ranks for none of these, even though it has published 57 Vibe Checks and two head-to-head reviews (Codex against Opus, and GPT-6 Sol against Opus 5.5).
| Search | US searches | Difficulty | Who ranks |
| --- | --- | --- | --- |
| claude vs chatgpt | 35,000 | 8 | Reddit, HackerNoon |
| is claude better than chatgpt | 12,000 | 1 | Reddit, Zapier |
| codex vs claude code | 11,000 | 11 | Towards AI, Duet |
| gemini vs chatgpt | 11,000 | 15 | Reddit, PCMag |
| chatgpt vs claude | 9,000 | 4 | Reddit, HackerNoon |
| claude code vs cursor | 5,000 | 8 | Reddit, blogs |
| claude vs gemini | 3,600 | 12 | Reddit, Gurusup |
| ai model comparison | 1,400 | 12 | Artificial Analysis |
| best llm for coding | 1,200 | 15 | Vellum, Reddit |
Ahrefs, US, November 2026. Below the forum threads, the results are blog posts and small review sites, several with a lower Domain Rating than Every's 77.
It’s the same pattern as “ai strategy consulting”, and Every has a team that uses these models all day and its own scores (the Vibe Checks report Reach Test ratings and a Senior Engineer score).
I’d start with one page that stays at one address and gets updated each time a model ships: “Claude vs ChatGPT: which one we use for each job”, at a comparison address in the Vibe Check section, such as `/vibe-check/claude-vs-chatgpt`. Each section answers one search:
| Section | Search it answers | US searches | Difficulty |
| --- | --- | --- | --- |
| Our verdict, by job | claude vs chatgpt | 35,000 | 8 |
| Is Claude better? | is claude better than chatgpt | 12,000 | 1 |
| Coding | claude vs chatgpt for coding | 600 | 22 |
| Writing | claude vs chatgpt for writing | 600 | 2 |
| What each costs | claude vs chatgpt pricing | 250 | n/a |
The sections of the Claude vs ChatGPT page and the search each one answers (Ahrefs, US, November 2026).
The verdict goes at the top, in two or three sentences, with the date it was last checked. Then the same template covers Gemini against ChatGPT, Codex against Claude Code and Claude Code against Cursor, each linking to the full Vibe Checks behind it.
### 2\. Claude Code: about 39,000 searches that lead to the camp
The course section above covers people looking for a Claude Code course. Far more people search for help with Claude Code itself, and Anthropic’s docs, Reddit and individual blogs answer them.
| Search | US searches | Difficulty | Who ranks |
| --- | --- | --- | --- |
| claude code pricing | 20,000 | 11 | Reddit, Anthropic |
| how to use claude code | 6,000 | 47 | Anthropic, Builder.io |
| claude code cost | 3,400 | 5 | Reddit, Anthropic |
| claude code hooks | 2,900 | 15 | Anthropic, Reddit |
| claude code tutorial | 2,800 | 17 | Anthropic, YouTube |
| claude code best practices | 1,800 | 0 | Hacker News |
| claude md | 1,700 | 18 | Anthropic, HumanLayer |
| claude code tips | 700 | 26 | GitHub, Anthropic |
Ahrefs, US, November 2026. Every teaches this in its camps and has written about how its six engineers work, but it ranks for none of these.
For “claude code pricing”, the first ordinary result is a Reddit post titled “Claude Pro at $20 is a trap for dev work”. First-hand numbers are what’s missing, and Every has them. I’d write one guide, “How Every’s engineers use Claude Code”, with these sections:
| Section | Search it answers | US searches | Difficulty |
| --- | --- | --- | --- |
| What it costs us each month | claude code pricing | 20,000 | 11 |
| What it costs us each month | claude code cost | 3,400 | 5 |
| The hooks and subagents we run | claude code hooks | 2,900 | 15 |
| Best practices from six engineers | claude code best practices | 1,800 | 0 |
| Our CLAUDE.md, line by line | claude md | 1,700 | 18 |
| Learn it live | claude code course | 1,000 | 59 |
| Best practices from six engineers | claude code tips | 700 | 26 |
The sections of the Claude Code guide and the search each one answers (Ahrefs, US, November 2026). The last section links to the permanent course page proposed above, with the next start date.
The compound engineering guide already shows this works: it ranks second for its own term and Google’s AI answers cite it.
### 3\. AI at work: the searches a consulting client makes first
Before a company hires consultants, someone there searches for how to start. These searches have low difficulty, and the pages that rank are vendor glossaries and research notes.
| Search | US searches | Difficulty | Who ranks |
| --- | --- | --- | --- |
| how to use ai | 5,500 | 8 | Ethan Mollick, Google |
| ai agents for business | 3,000 | 15 | BCG, Lindy |
| ai adoption | 2,800 | 19 | Fortinet, the Fed |
| ai transformation | 2,400 | 24 | CGI, IBM |
| ai readiness assessment | 1,800 | 12 | Microsoft, Cisco |
| ai adoption framework | 1,100 | 0 | deepset, Xebia |
| ai maturity model | 800 | 28 | Gartner, CMU |
| ai policy template | 600 | 11 | NTEN, Responsible AI |
Ahrefs, US, November 2026. Every's guide "The eight levels of AI adoption" covers this ground but gets about 18 visits a month, for "ai levels".
I’d turn the eight levels guide into the page for “ai adoption framework” and “ai maturity model”: the same framework, with a title that names the search, a short definition at the top, and one client story per level. Every’s “executive’s guide to implementing AI” can answer “ai transformation” the same way. Both should end on the consulting page.
### Free tools and pages built from a pattern
Searches with a tool in them (“assessment”, “template”, “generator”) come from someone who wants to do a job now, and a tool page can end at the product that does the rest. These are the ones Every can build from what it already has:
| Search | US searches | Difficulty | Who ranks | Built from |
| --- | --- | --- | --- | --- |
| ai readiness assessment | 1,800 | 12 | Microsoft, Cisco | The eight levels guide |
| ai policy template | 600 | 11 | NTEN, Responsible AI | The consulting team |
| linkedin post generator | 500 | 12 | Hootsuite, Postiv | Spiral |
Ahrefs, US, November 2026. Each tool is built from something Every already owns. The readiness assessment would also target "ai readiness" (1,300, difficulty 17), and the policy template "ai acceptable use policy template" (200, difficulty 3).
The AI readiness assessment would be ten questions that place a company on Every’s eight levels, then a results page that explains the level and what to do next, with a link to book a consulting call. Microsoft, Cisco and PwC rank with their own versions.
The AI policy template is the policy a company needs before its staff use AI at work, as a page and a download. It’s an easy first step toward the consulting team.
The LinkedIn post generator would be a free, limited version of Spiral that writes one post from a pasted idea, with a link to the full app. Hootsuite and Grammarly rank with free tools like this.
The bigger pattern is one page per model. The Rundown’s AI tool directory has a page for each tool, and its Claude page alone gets an estimated 88,800 of its 151,600 visits. Every can do the same with models: one page per model that holds Every’s verdict, the jobs it’s best at and links to each Vibe Check. Model names get real searches: “claude opus 5.5” has 2,600 US searches a month, “gemini 4 argon” 1,400 and “claude sonnet 5.5” 900 (Ahrefs). These pages are the model family hubs from the model review section, one level down.
## AI search: the answers rarely cite Every
AI answers cite Every for one of 10 questions a buyer of its products might ask. Crawler access isn’t the reason, so most of the fix is in what the pages say.
### What AI crawlers get
GPTBot, OAI-SearchBot, ClaudeBot and PerplexityBot user agents get the same pages a browser gets, and llms.txt and Markdown copies of the posts exist. The gaps are the Vibe Check verdicts and the consulting page.
1. Fix
/vibe-check/gpt-5-6-sol
Every user agent gets the 376-word intro, then "Start your free trial to read the full Vibe Check". That link redirects back to the same page.
2. Add
/consulting.md
Returns the HTML page, not Markdown.
3. Fix
/consulting
Has only the site's Organization and WebSite markup, with no structured data that says Every offers consulting.
4. Fix
/llms.txt
About 22 KB of plain text. Lists the consulting page in one line, near the end.
5. Keep
/robots.txt
Allows GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-User, PerplexityBot and Google-Extended, with Content-Signal search=yes, ai-input=yes, ai-train=yes.
6. Keep
/llms-full.txt
The 47 latest posts as Markdown.
7. Keep
/guides/compound-engineering.md
The full guide as Markdown, about 9,500 words.
What every.to serves to AI crawlers, 4 November 2026.
The Vibe Checks matter most. The verdict an AI answer would quote, “which model gets which job”, is behind the trial. That’s a fair business choice for the full review. But the verdict alone is the part that gets cited, and it’s what brings a reader to the trial.
Today
…the full Vibe Check includes our Reach Test ratings,
benchmark results, screenshots, pricing analysis, and
exact guidance on which model should get which job.
\[Start your free trial to read the full Vibe Check\]
→ /vibe-check/gpt-5-6
→ 308 to /vibe-check/gpt-5-6-sol (this page)
With a free verdict
Our verdict
Use GPT-5.6 Sol for \[the jobs it won\].
Use Fable for \[the jobs it won\]. Checked \[date\].
…the full Vibe Check includes our Reach Test ratings…
\[Start your free trial to read the full Vibe Check\]
What a crawler, or a reader without a subscription, gets at the end of the free part of a Vibe Check, today and with a free verdict.
The consulting page needs a Markdown copy, service markup and a better place in llms.txt. The guides have no Article markup either, unlike the posts.
### Who AI answers cite
| Question | Every | The Rundown | Cited instead |
| --- | --- | --- | --- |
| Best AI consulting firms for companies adopting AI | Named in AI Mode | No | McKinsey, Accenture |
| Best AI email assistant | Cora: no | No | Not checked |
| Best AI newsletter to keep up with AI | No | Cited by 3 of 3 | therundown.ai |
| Best AI writing tool for social media posts | Spiral: no | No | Not checked |
| Best course to learn Claude Code | No | No | Anthropic, DataCamp |
| Best dictation app for Mac | Monologue named in AI Mode | No | Not checked |
| Claude vs ChatGPT: which is better for writing | No | No | Zapier, Engadget |
| How do I train my employees to use AI | No | No | No links |
| What is compound engineering | Cited by 2 of 3 | No | every.to |
| Which AI model is best for coding right now | No | No | BenchLM, Builtin |
AI answers on 4 November 2026 from ChatGPT search, Google AI Mode and Google's AI Overviews (DataForSEO, US), for every.to, The Rundown AI and Every's app sites. Cited means the answer links to the site; named means it names it without a link. The last column lists sites ChatGPT and AI Mode cited instead.
The one question that cites Every is the one where it owns the term. The Rundown is also cited for one: “best AI newsletter”, by all three engines, where Every isn’t even mentioned. The cited pages for the other questions have something Every’s pages don’t:
For consulting, the answers cite lists of AI consulting firms (TechRadar, Forrester and several small list sites) and the big firms’ own pages. Every is named once, in AI Mode, without a link. Getting into those lists is the same job as in the consulting section.
For courses, they cite course pages with a syllabus: Anthropic’s academy, DataCamp, Coursera, Scrimba and DeepLearning.AI. Every has no permanent course page, and its courses address redirects to the events page.
For the model questions, they cite comparison pages and leaderboards with a date and a verdict near the top. Every’s verdicts are behind the trial, and it has no comparison pages.
For the newsletter question, The Rundown’s homepage is also the first ordinary result for “ai newsletter” (2,100 US searches a month). Every’s newsletter page ranks seventh for “ai newsletters”.
Most of the fixes below are pages from the new content section, written so the answer an AI would quote sits in the first lines.
**The fix:**
- Put a free verdict block above the trial on every Vibe Check: two or three sentences on which model to use for which job, and the date.
- Publish the comparison pages, the Claude Code guide and the permanent course page, each with the direct answer in its first paragraph.
- Give the consulting page a Markdown copy like the posts have, and add structured data for the consulting service (a `ProfessionalService` with the services it offers).
- Rewrite the top of the newsletter page to say what Every sends, how often and who reads it, and pitch Every to the “best AI newsletters” lists that rank, such as Zapier’s.
## Three technical fixes, closest to revenue first
The crawl found three technical issues. I’ve put them in order of how close each one is to revenue, so the first one isn’t the one that affects the most pages.
### 1\. The guides send their canonical tag too late
A canonical tag tells Google which address is the main copy of a page. [Google only reads it in the head](https://developers.google.com/search/docs/crawling-indexing/consolidate-duplicate-urls) of the HTML. On every guide I tested, the tag arrives in the body, so Google ignores it and picks a main copy on its own. Ordinary posts are fine (22 of 22 tested).
Guide page
…
…
← too late
Ordinary post
…
…
Where the canonical tag sits in the HTML of a guide page, compared with an ordinary post.
Guides are the evergreen pages most likely to rank and to link to products, so they’re the pages where this matters most.
**The fix:** output the canonical tag in the head of the guide template, as the post template already does. It should hold the full address of the page itself, such as `https://every.to/guides/compound-engineering`. Then check one guide in Search Console’s URL Inspection, where the “User-declared canonical” should show that address.
### 2\. Compound Engineering has four pages
Four every.to pages cover the same idea. The guide ranks second or third for “compound engineering” (1,600 US searches a month). The other three split the links that could push it to first, and Google has to work out which one to show.
1. Keep
/guides/compound-engineering
The page that ranks. Move anything newer from the other pages into it.
2. 301
/source-code/compound-engineering-the-definitive-guide
redirects to /guides/compound-engineering
A second guide on the same subject.
3. Keep
/chain-of-thought/compound-engineering-how-every-codes-with-agents
The original essay. Keep it, and link to the guide in its first paragraph.
4. 301
/source-code/compound-engineering-how-every-codes-with-agents-af3a1bae-…
redirects to /chain-of-thought/compound-engineering-how-every-codes-with-agents
A copy of the essay at a second address. Fix 3 covers these.
The four Compound Engineering pages and what I'd do with each. All four return the page today, and each names itself as the canonical.
**The fix:** one page for each search. Merge the definitive guide into the guide and redirect it there with a permanent (301) redirect. Keep the original essay, since it has its own links and history, and have it send readers on to the guide. Then update the internal links and the sitemap so they only list the pages that stay.
### 3\. 72 posts are published twice
Each of these posts has a twin: the same address with a long ID on the end. Both addresses return the full post, both name themselves as the canonical, and both are in the sitemap. So Google sees two equal pages and splits the links between them.
Today
/source-code/stop-coding-and-start-planning
200 OK
canonical: itself
/source-code/stop-coding-and-start-planning-be0b4fd1-…
200 OK
canonical: itself
After the fix
/source-code/stop-coding-and-start-planning
200 OK
canonical: itself
/source-code/stop-coding-and-start-planning-be0b4fd1-…
301 Moved Permanently
→ /source-code/stop-coding-and-start-planning
What a visitor or Googlebot gets from each address, today and after the fix.
**The fix:** a permanent (301) redirect from every ID address to the clean one, set up as one pattern rule on the server or CDN. The sitemap and internal links should list only the clean addresses. If the CMS needs the ID addresses for previews, they can stay, as long as each one redirects for everyone else. Some republished posts also have “-rerun” copies. For those, I’d update the original post and its date, or point the rerun’s canonical tag at the original.
## Ship it all today, then let it settle
Teardowns like this usually end with a four-week plan, with the links in week one, the consulting pages in week two and so on. With coding agents, I’d start all of it on the same day.
Each fix above is a separate job with a clear end, so ten agents can work on them at once:
1. Product links
Change the newsletter template so each product mention links to the page that sells it, with descriptive link text. Fix the monologue.co links.
2. Guide canonicals
Make guides send the canonical in the head, like posts already do.
3. Consulting hub
Draft strategy, implementation and team training pages from the existing consulting page and client stories, for the team to edit.
4. Course page
One permanent Claude Code course page that past and future cohorts point to, plus a courses hub.
5. Model reviews
Model-first title tags on Vibe Checks, and a hub page per model family.
6. Duplicates
Redirect the ID-suffixed copies and merge the Compound Engineering versions.
7. Comparison pages
Draft Claude vs ChatGPT from the existing Vibe Checks, with the verdict on top, then the same template for Gemini, Codex and Cursor.
8. Claude Code guide
Draft the guide from the engineers' own CLAUDE.md files, hooks and monthly spend, linking to the course page.
9. Free tools
Build the AI readiness assessment from the eight levels guide, and publish the AI policy template.
10. AI search
Add a free verdict block to each Vibe Check, a Markdown copy and service markup for the consulting page, and a clearer top to the newsletter page.
Someone at Every still needs to review the consulting copy, the course page, the client stories and each verdict, because they know the business and the models. The link, canonical, title and redirect changes need only a quick check.
Then I’d spend 30 days checking the results:
1. Day 0: Ask Google to recrawl
Submit every new and changed URL to Index My Site, so Google recrawls them now instead of on its own schedule.
2. Day 3: Check indexing
Confirm the new pages are indexed and the guides now send the canonical in the head.
3. Day 7: First impressions
Look for impressions on "ai strategy consulting", "claude code course" and "claude vs chatgpt" in Search Console.
4. Day 14: Product clicks
Check clicks from every.to links to the product pages in analytics. Those clicks are the first sign of new revenue.
5. Day 30: Enquiries, sign-ups and AI answers
Compare consulting enquiries and course sign-ups from search with the month before, and ask the AI answers the same 10 questions again to see which ones now cite Every.
## Get the changes seen quickly
Almost everything above changes pages that Google has already crawled. Google comes back on its own schedule, which for older posts can be weeks. I’d submit every changed and new URL to [Index My Site](https://indexmysite.app/) as it ships. That matters most for the retitled reviews, the merged Compound Engineering guide and the redirected duplicates, where Google needs to see the change before anything moves.
## How I ran this
Everything in this teardown came from SEO Skill, run by an agent. You can run the same audit on your own site.
**1\. Connect the hosted MCP server.** It’s free. You sign in with Google, and there’s no password and nothing to install. In Claude Code:
```sh
claude mcp add --transport http seoskill https://seoskill.dev/mcp
```
Codex, Cursor and Claude on the web work too. The [MCP guide](https://seoskill.dev/docs/mcp) has the steps for each one.
**2\. Add a research key.** On your [account page](https://seoskill.dev/account), save an Ahrefs or DataForSEO key. The key gives your agent search volumes, rankings and results pages, like the tables above. With DataForSEO it can also ask ChatGPT and Google’s AI answers your questions and check who they cite.
**3\. Install the CLI for crawling.** The crawl runs on your machine, so it can read any site:
```sh
npm i -g seo
seo login
```
**4\. Ask your agent.** This is close to the prompt I used:
```text
Audit example.com and every site we own.
Crawl each one with seo crawl. Find where our sites
link to each other and which pages those links hit.
Use seo_research to find the searches our buyers
make, who ranks for them and whether we can compete.
Find the searches our competitors rank for and we
don't, and group them into pages we should write.
Ask ai_answers 10 buyer questions for us and our
top competitor, and say who gets cited.
Rank every fix by how close it sits to revenue.
```
On your own site, the hosted server adds the part I couldn’t see from the outside: your real Search Console clicks and queries. With those, you get measured numbers where this post has estimates.
If you have more than one product, crawl the main site too and look at where it links to the others. The cheapest fixes in this teardown came from those links.
If you’re at Every, I’m happy to talk any of this through. And if you’d like the same teardown for your company, [SEO Skill](https://seoskill.dev/) will run it for you.