--- title: "How I'd grow Ben's Bites with SEO | SEO Skill" description: "An SEO teardown of Ben's Bites, Ben Tossell's AI newsletter. Where its Google traffic went, the closed catalog, the old domains and the fixes I'd ship." canonical: "https://seoskill.dev/blog/how-id-grow-bens-bites" language: "en" --- # How I'd grow Ben's Bites Ben's Bites has more than 118,000 subscribers and about 890 visits a month from Google. Most of those visits land on a catalog that now returns a 404. This is how I'd turn search back into a steady source of new readers. 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 ![Ben's Bites' red mascot holds up a rescued index card in an abandoned library of open, empty card catalogue drawers.](https://seoskill.dev/blog/covers/how-id-grow-bens-bites.webp) [Ben’s Bites](https://www.bensbites.com/) is Ben Tossell’s AI newsletter. It goes out on Tuesdays and Thursdays to more than 118,000 subscribers, and it makes money from sponsors: one main ad and one tool placement in each send. Its sponsor page prices the main ad at about a $48 CPM, so each new subscriber adds revenue to every send for as long as they stay. For a newsletter, that’s the job search has to do: bring in readers who haven’t heard of it yet. Ben is also changing the newsletter right now. In [Finding my thing](https://www.bensbites.com/p/if-you-feel-behind-right-now-dont) he says the format, topics and posting days will all vary until the end of the year (“I’m in experimental mode”). Search is the channel that keeps bringing in new readers whatever the format, as long as the pages they land on still exist. ## Most of Ben’s Bites’ search traffic lands on pages that no longer exist Ahrefs estimates Ben’s Bites gets about 885 visits a month from Google across all its domains. This map shows where they land. - Catalog tutorials363 - Searches for Ben's Bites by name200 - Newsletter posts found by topic133 - Catalog tool pages88 - Catalog pages found by name73 - Catalog articles28 Estimated Google visits to bensbites.com and its subdomains by kind of page, November 2026 (Ahrefs top pages). Blue pages bring in readers who weren't searching for Ben's Bites. Red pages now return a 404. About 62% of those visits land on catalog.bensbites.com, the tutorials and tools library behind the old Pro membership. Every page I tried there now returns a “Page not found” error. Only about 15% reach a newsletter post from someone who wasn’t already looking for Ben’s Bites. The catalog wasn’t a side project in search terms. Over the last year it was the part that grew, while the newsletter posts fell. | Month | Catalog | Newsletter | Note | | --- | --- | --- | --- | | May 25 | 111 | 314 | | | Jun 25 | 114 | 896 | | | Jul 25 | 131 | 1,449 | | | Aug 25 | 126 | 1,508 | | | Sep 25 | 103 | 1,454 | | | Oct 25 | 74 | 1,191 | | | Nov 25 | 84 | 864 | | | Dec 25 | 107 | 785 | | | Jan 26 | 78 | 899 | | | Feb 26 | 142 | 1,027 | | | Mar 26 | 155 | 1,084 | | | Apr 26 | 324 | 852 | | | May 26 | 948 | 802 | The catalog's best month, about 950 visits | | Jun 26 | 521 | 754 | | | Jul 26 | 328 | 683 | | | Aug 26 | 469 | 449 | | | Sep 26 | 620 | 274 | | | Oct 26 | 805 | 294 | | Estimated monthly Google visits to the catalog and to the newsletter on www.bensbites.com (Ahrefs). The catalog now returns a 404 on every page. Ahrefs still credited the catalog with about 805 visits in October. Those estimates come from where the pages ranked, so they lag behind a shutdown. Ahrefs’ 1 November data still has the ChatGPT tutorial first for “chatgpt interface”, but its latest results page for that search no longer shows it. The traffic is going, and it goes with the pages. I checked all 1,020 catalog addresses the Wayback Machine has saved with SEO Skill. 1,016 return a 404, and the four that don’t are robots.txt and other system files. That includes 433 tutorials, 201 articles and 122 tool pages. ## New readers: the bigger AI newsletters get them from evergreen pages These are Ahrefs’ estimates of this month’s Google traffic for the AI newsletters people compare with Ben’s Bites. 1. therundown.aiAbout 93% on its AI tools directory: 159,650 visits/month 2. tldr.techMostly people searching for "tldr": 39,585 visits/month 3. latent.space: 5,453 visits/month 4. superhuman.ai: 3,985 visits/month 5. bensbites.com: 885 visits/month 6. alphasignal.ai: 866 visits/month 7. theneurondaily.com: 816 visits/month Estimated monthly organic visits, November 2026 (Ahrefs, all subdomains). 148,197 of The Rundown's visits land on its AI tools directory. The Rundown’s lead isn’t its daily issue. About 148,000 of its 160,000 visits land on its AI tools directory, one page per tool. Ben’s Bites had the same kind of asset: 122 tool pages in the catalog, one of which ranked second for “chatbot arena”. They all return a 404 now. I wouldn’t copy The Rundown’s directory. Most of its traffic is people typing “claude ai” into Google, who wanted Claude, not a newsletter. The part worth copying is that each evergreen page brings in readers every day without an issue going out. The searches for the category itself are small, and Ben’s Bites isn’t in them. “ai newsletter” gets about 2,100 US searches a month (Ahrefs), and these are the pages Google shows. ai newsletter 2,100 searches a month 1. 2therundown.aiNewsletter homepageDR 72 2. 3tldr.techNewsletter pageDR 72 3. 4reddit.comForum threadDR 95 4. 5medium.comList postDR 94 5. 6github.comList of newslettersDR 97 6. 7ibm.comNewsletter sign-upDR 93 7. 9venngage.comNewsletter generatorDR 84 8. 10zapier.comList postDR 91 9. ×bensbites.comNot ranking in these results DR is [Domain Rating by Ahrefs](https://ahrefs.com/website-authority-checker). Organic results for "ai newsletter" (Ahrefs, US). Above them, Google's AI Overview cites The Rundown, TLDR, Superhuman and five list posts. None of them is Ben's Bites. Most of this results page is lists, so the way in is the lists. Ben’s Bites is already in Zapier’s (tenth here and fifth for “best ai newsletter”), Leland’s and Clickittech’s. daily.dev’s list still links to bensbites.co. It’s missing from the GitHub list in sixth place, the DLabs and Lymlyt lists the AI Overview cites, and a Substack post ranking the 100 best AI newsletters on Substack. Each one is an email to the author. Those lists and AI answers send people to the homepage, and the homepage’s search result doesn’t say what Ben’s Bites is. Its title is Substack’s default, “Ben’s Bites | Substack”, and the description ends in Substack’s stock line. Today Ben's Biteshttps://www.bensbites.com Ben's Bites | Substack What I'm learning building with agents (plus tools worth trying). Click to read Ben's Bites, a Substack publication with hundreds of thousands of subscribers. Title 22 characters. Description 158 characters. After the fix Ben's Biteshttps://www.bensbites.com Ben's Bites: an AI newsletter for builders Ben Tossell on what he's learning building with AI agents, plus the tools he's using. Free, read by more than 118,000 subscribers. Title 42 characters. Description 130 characters. The homepage as it shows in Google today, and with a title and description written for someone who hasn't heard of Ben's Bites. Both are set in Substack's SEO settings. The new title names the topic and who it’s for, so it can match “ai newsletter” searches, not only searches for the name. ## Tutorials: the pages that brought in readers, and where they could live now The catalog’s tutorials taught ChatGPT, Claude and other tools to people at work. Ben’s writing has moved on to building with agents (“I can’t code. I build with agents”, says his cookbook). The searches have moved the same way. | Search | US searches | Difficulty | Who ranks | | --- | --- | --- | --- | | how to use claude code | 6,000 | 47 | Anthropic, Builder.io, Reddit, Product Talk | | claude code tutorial | 2,800 | 17 | Anthropic, YouTube, a Substack guide, a free course site | | how to use claude cowork | 2,100 | 1 | A Substack newsletter, YouTube, Reddit | | how to vibe code | 900 | 12 | Reddit, Microsoft, Wikipedia, roadmap.sh | | how to build an app with ai | 600 | 31 | Reddit, app builders, Zapier | | how to code with ai | 150 | 9 | Reddit, a personal blog, Medium | Ahrefs, US. Ben's Bites has no page that ranks for any search containing "claude code" or "vibe". Two of these results pages have another newsletter's Substack post on page one. “how to use claude cowork” is the clearest one. It gets 2,100 searches a month at a difficulty of 1, and the second result is a post on another newsletter’s Substack. how to use claude cowork 2,100 searches a month 1. 2thesignal.substack.comNewsletter postDR 94 2. 4youtube.comVideoDR 99 3. 6reddit.comForum threadDR 95 4. 7youtube.comVideoDR 99 5. 9claudecowork.imUnofficial guideDR 18 6. 10productmarketer.substack.comNewsletter postDR 94 7. ×bensbites.comNot ranking in these results DR is [Domain Rating by Ahrefs](https://ahrefs.com/website-authority-checker). Organic results for "how to use claude cowork" (Ahrefs, US). Two of them are posts on newsletters' own Substacks. Ben’s Bites has written about Cowork, in a news issue called “New home for Cowork”. A news issue answers “what happened this week”. The searcher wants “how do I set it up”, and that needs its own page with a title that says so. Ben already has the start of that page type. His cookbook has a 4,700-word “Become Technical in 1 Day” guide and “How I Code with Agents, Without Being ‘Technical’”. The cookbook gets no measurable search traffic yet (Ahrefs), and six sites link to it. I’d put these guides where the authority already is: on bensbites.com, as posts next to the newsletter. That’s the domain 4,000 sites link to (Ahrefs). The cookbook is a separate subdomain with six. Each guide answers one search, has a title that uses the words people search, and gets a link from the issue that announces it. The issues keep their inbox titles. ## Ben’s best X posts could be his best search pages Ben has more than 200,000 followers on X, and his best posts there are guides on the subjects people search for. They live on x.com, where they bring in a spike of readers and then stop. His X Article “How I code with agents, without being ‘technical’” went out on New Year’s Eve 2025 and has more than 4 million views and 24,000 bookmarks (November 2026). 28 sites link to it on x.com (Ahrefs). A week later he republished it on the newsletter, where 12 sites link to it, and there’s a third copy on the cookbook. None of the three gets measurable search traffic. His follow-up, “What you need to know to build with coding agents, if you’re not technical”, got about 580,000 views in January. That’s the “Become Technical in 1 Day” guide on the cookbook. Each of its sections answers a search beginners make, and the cookbook also has four short glossary pages that its homepage doesn’t link to. | Search | US searches | Difficulty | Ben has it as | | --- | --- | --- | --- | | claude skills | 21,000 | 70 | The "Skills" section of the guide | | what is docker | 14,000 | 73 | A 237-word glossary page | | what is a vps | 3,500 | 39 | A 238-word glossary page | | what is bash | 2,800 | 7 | Part of the "Commands" section | | agents.md | 1,400 | 53 | The "Markdown Files & agents.md" section | | coding agents | 1,100 | 25 | The "What Coding Agents Are" section | | what is a terminal | 1,100 | 5 | The "Terminals" section | | system prompt | 700 | 9 | The "System Prompts" section | | what is an api token | 350 | 0 | A 149-word glossary page | Sections of Become Technical in 1 Day and the cookbook's glossary pages, with the search each one could answer (Ahrefs, US). None of these pages ranks for its search today. “what is a terminal” and “what is bash” have difficulties of 5 and 7. Reddit threads, Wikipedia, Apple’s manual and programming tutorials rank for them now, all written for people who already code. Ben writes for people who don’t code but work with agents every day, and nobody on page one writes for them. I’d split the guide into one post per question, each with a title that uses the words people search, and keep the full guide as the hub that links to them. Then I’d use the same pattern for every post that does well on X: 1. Post on X as now. That’s where the spike comes from. 2. Within a week, publish the long version on bensbites.com, one post per question it answers. Substack can publish a post to the web without emailing it. 3. Send a short newsletter issue that links to the guide, instead of another full copy, so Google has one page to rank. 4. Reply to the X post with a link to the guide, so people who find the post later end up on Ben’s own site. The views on an X post come in the first few days. A page that ranks for “what is a terminal” is in front of about 1,100 searchers a month (Ahrefs, US), every month, and they’re the people Ben writes for. The New content section below turns this guide and two more of Ben’s posts into pages, each with the search it targets. ## One domain instead of seven Ben’s Bites is spread across a lot of addresses: the newsletter on [www.bensbites.com](http://www.bensbites.com/), the catalog, the cookbook, a sessions site, a partner page, and the old bensbites.co and news.bensbites.com. Search engines treat each subdomain as its own site, so every one starts from zero, needs its own Search Console property and has to earn its own links. The newsletter has the authority. Ahrefs counts about 4,000 sites linking to [www.bensbites.com](http://www.bensbites.com/), against six for the cookbook. Everything new should go there, and every old address should end up there. 1. Keep www.bensbites.com The one site. Newsletter issues, guides and restored tutorials all live here, in one sitemap and one Search Console property. 2. 301 cookbook.bensbites.com redirects to www.bensbites.com/p/… Republish each guide as a post, and redirect each page to its copy. 3. 301 catalog.bensbites.com redirects to www.bensbites.com/p/… 95 articles already have copies. Restore the tutorials that ranked as posts, and return 410 for the rest. 4. 301 sessions.bensbites.com redirects to www.bensbites.com/p/… The newsletter already has "Ben's session" posts. Put the session write-ups there. 5. 301 bensbites.co and news.bensbites.com redirects to the matching post, or 410 Fix 2 below. 6. Leave partner.bensbites.com A sponsor page nobody needs to find in search. Every Ben's Bites address I found, and what I'd do with each, so everything ends up on one site. Substack decides how far this can go. It needs [www.bensbites.com](http://www.bensbites.com/) pointed straight at its servers, with Cloudflare’s proxy off ([Substack’s setup guide](https://support.substack.com/hc/en-us/articles/360051222571-How-do-I-set-up-my-custom-domain-on-Substack)), so Ben can’t serve a folder like /guides from anything else on that domain. That’s probably why the cookbook and sessions sites ended up on subdomains. Inside Substack, the guides go where the posts go, under /p/. For search, that’s fine. Every post’s full text is in the HTML, and each post has a canonical tag and NewsArticle structured data. What Substack doesn’t offer is control over templates and redirects, an llms.txt or a Markdown copy of each post (the AI search section below has what each AI crawler gets). Those gaps are much smaller than splitting the guides across another domain, so I’d keep everything on Substack. If Ben later wants that control, the move is a site he runs himself on [www.bensbites.com](http://www.bensbites.com/), with the same /p/ addresses, rather than another subdomain. ## Ben’s old issues link to 566 dead addresses Ben’s Bites has used a lot of addresses over four years: bensbites.co, a news site on news.bensbites.com, a catalog that lived at bensbites.com before it moved to its own subdomain, a beehiiv newsletter, and now Substack, the cookbook, Ben’s sessions site and a sponsor page. Old issues still link to all of them. I pulled every link out of the body of all 270 posts on the newsletter and checked where each one ends up today. 1. Catalog tutorialsAll 185 return a 404: 185 links 2. Catalog articles85 return a 404. 41 reach the newsletter copy: 126 links 3. Workshop recordingsAll return a 404: 72 links 4. Topic pagesAll return a 404: 66 links 5. Become a member linksLand on the newsletter homepage: 65 links 6. Pro sign-up and pricingAll return a 404: 62 links 7. news.bensbites.com tool listsRedirect to the newsletter homepage: 62 links 8. Courses and tool pagesAll return a 404: 41 links 9. Everything elsePolls, forms, the cookbook, Design Words and a few more: 59 links Links from the 270 newsletter posts to Ben's other addresses, by destination (each address checked on 4 November 2026). 566 of the 738 links end on a 404, and 127 land on the newsletter homepage. These links point at bensbites.com addresses from when the catalog lived there. bensbites.com now forwards every address to the newsletter with the path kept, and Substack has no page at that path, so a reader who clicks “Build your own ChatGPT in Slack” in an old issue gets a 404. The newer sites get almost nothing. Across all 270 posts I found one link to the cookbook and three to Ben’s sessions site, which is another reason to put those pages on the newsletter’s own domain. **The fix:** add the old bensbites.com catalog addresses to the same redirect list as the catalog subdomain in fix 1 below. bensbites.com already runs through Cloudflare, so one rule there catches the 566 dead links in every old issue at once, without editing a post. Then link each new issue that teaches something to the matching guide, with link text that names it. ## New content: the pages the bigger newsletters rank with Apart from searches for their own names, the bigger AI newsletters get their search traffic from three kinds of page, and Ben’s Bites doesn’t have any of them yet. | Search | US searches | Difficulty | Who ranks | | --- | --- | --- | --- | | notion ai | 37,000 | 19 | Rundown tool page | | websim | 24,000 | 1 | Latent Space wiki | | claude dashboard | 2,600 | 1 | Neuron news post | | ai newsletter | 2,100 | 39 | Rundown, TLDR | | chatgpt picture generator | 1,900 | 49 | Superhuman guide | | news ai | 1,500 | 23 | Superhuman home | | best ai newsletter | 1,300 | 68 | Rundown, TLDR | | agentic harness | 450 | 45 | Latent Space essay | | data engineering news | 450 | 6 | TLDR Data page | | what is cline | 450 | 0 | Latent Space post | | ai newsletters | 350 | 4 | TLDR, Superhuman | | difference between grok and chatgpt | 350 | 5 | Superhuman guide | Searches where a bigger AI newsletter ranks in the top 20 and Ben's Bites has no page, with the competitor's page that ranks (Ahrefs, US, November 2026). The first kind is a page that says what the newsletter is. “ai newsletter” and “best ai newsletter” go to a homepage or newsletter page whose title names the category. That’s the homepage title fix above. TLDR also has a page for each of its newsletters, and its data one is 10th for “data engineering news”. The second is evergreen guides and comparisons. Superhuman’s “Grok AI vs ChatGPT” (about 7,200 words) ranks first for “difference between grok and chatgpt”, and its ChatGPT image generation guide (about 7,100 words) is 11th for “chatgpt picture generator”. A team writes them for general readers. Ben can write the same kind of page from tools he has tested himself. The third is a page per tool or company, built from what the newsletter already has. The Rundown’s directory gives every tool the same sections: what it is, who it’s for, features and how it works. Latent Space builds a wiki page per company and person from its own archive. Its WebSim page is a timeline by year plus links to every Latent Space piece on it, and it gets about 870 visits a month from “websim” (Ahrefs). Ben’s Bites has four years of issues about the same tools and people. The pages below come from that archive and from Ben’s best X posts. Each one gets a title that uses the search, a two-sentence answer at the top, and a link from the issue that announces it. ### Coding agent comparisons, from the Vibe Coding Deathmatch Ben’s most linked post on X is the “Vibe Coding Deathmatch” from March 2025, a video of Bolt, Replit, Cursor and Windsurf building the same project (Replit won). Ahrefs counts 1,731 sites linking to it, more than any other post of Ben’s. The newsletter write-up, “AI coding tools: Head-to-head”, ranks for nothing in Ahrefs, and neither does “What AI coding tool should you use?”. These searches have grown since then, and their difficulty is still low. “cursor vs claude code” gets 6,100 searches a month at a difficulty of 5, and a post on someone’s Substack ranks sixth. Another Substack post is sixth for “claude code vs codex”. I’d make one hub and one page per pair, each from Ben’s own test, with the verdict in the first two sentences. | Target search | US searches | Difficulty | Ranks first | | --- | --- | --- | --- | | cursor vs claude code | 6,100 | 5 | ksred.com | | claude code vs codex | 4,800 | 13 | Reddit | | best ai coding tools | 2,900 | 9 | Zapier | | best vibe coding tools | 2,300 | 9 | Zite | | replit vs lovable | 1,200 | 2 | Superblocks | | codex vs cursor | 1,100 | 1 | Medium | Searches for the proposed comparison pages (Ahrefs, US). Ben's Bites ranks for none of them today. The pages, each on bensbites.com at /p/ plus its search: 1. The hub, “The best vibe coding tools, tested head to head”, at /p/best-vibe-coding-tools. 2. “Cursor vs Claude Code, if you don’t code”, at /p/cursor-vs-claude-code. 3. “Claude Code vs Codex: which I use for what”, at /p/claude-code-vs-codex. 4. “The best AI coding tools for people who don’t code”, at /p/best-ai-coding-tools. 5. “Replit vs Lovable” and “Codex vs Cursor”, at /p/replit-vs-lovable and /p/codex-vs-cursor. Each page covers what the two tools are, the same project built in both, what each got wrong, the cost, and which one Ben would give a beginner. Ben works at Factory (“I use Droid because I work at Factory”, he writes in the agents article), so each page should say that at the top and include Droid in the test. A stake readers can see reads better than one they find out about later. ### “How I code with agents” as a page that ranks The newsletter copy keeps the title from X, which no one types into Google. I’d keep the address, since 12 sites link to it, and change the title to “How to code with AI agents when you’re not technical”, with a short answer above the first section. Then each section links to the page that answers its question in full. | Section | Search | US searches | Difficulty | | --- | --- | --- | --- | | Beyond "vibe coding" | what is vibe coding | 50,000 | 42 | | What I've been learning: VPS | what is a vps | 3,500 | 39 | | What I've been learning: bash commands | what is bash | 2,800 | 7 | | CLIs over MCPs | what is a cli | 1,500 | 26 | | My agents.md setup | agents.md | 1,400 | 53 | | What I've been learning: skills | agent skills | 1,400 | 42 | | Just pick one | coding agents | 1,100 | 25 | | How I actually work | how to vibe code | 900 | 12 | | How I actually work | how to code with ai | 150 | 9 | Sections of the agents article and the search each one answers (Ahrefs, US). The article would target the first two searches. The rest link to their own pages. The biggest one already has a post. Ben published “What is ‘vibe coding’?” in February 2025, the month the term took off. “what is vibe coding” now gets 50,000 searches a month (Ahrefs, US), and the post ranks for none of them. Page one is IBM, Reddit, Wikipedia, Cloudflare and Replit, all explaining it from the outside. Ben does it every day and has opinions on the name (“Vibe coding is old now”). I’d update that post with a two-sentence definition at the top, a section on how to vibe code and links to the comparisons. At a difficulty of 42 it’s a slower win than the comparisons. ### Become Technical in 1 Day, one page per question The table in the X section maps the guide’s sections. As pages, they make a glossary for people who build with agents but don’t code, at /p/ plus the search. | Target search | US searches | Difficulty | | --- | --- | --- | | what is an mcp server | 6,300 | 47 | | what is a vps | 3,500 | 39 | | what is bash | 2,800 | 7 | | context window | 2,100 | 36 | | what is a cli | 1,500 | 26 | | what is a terminal | 1,100 | 5 | | what is localhost | 900 | 39 | | system prompt | 700 | 9 | | what is an api token | 350 | 0 | | what are environment variables | 300 | 24 | Proposed glossary pages, from the guide and the cookbook's glossary (Ahrefs, US). Each is titled as the question, such as "What is a terminal? A guide for people who don't code" at /p/what-is-a-terminal. Every glossary page gets the same four sections: the answer in two sentences, why it matters when you work with an agent, the one command or setting you’ll actually use, and a link back to the hub. Once the first few are written, each new term is a short job. ### Free tools and a page per tool Ben already builds free tools. His personal site has a token usage tracker (10,011 views on X), a live view of which models OpenRouter’s top apps use (15,395 views) and “50 years of devices” (31,202 views). Ahrefs credits the site with about 10 visits a month from search, all for Ben’s name. Substack can’t host an interactive page, so the tools stay on his personal site. What’s missing is a post on bensbites.com that answers the search and links to the tool. The same goes for a page per coding agent, in the Latent Space pattern: what it is, what Ben built with it, his verdict, and every issue that covered it. | Search | US searches | Difficulty | Ben has | | --- | --- | --- | --- | | what is claude code | 11,000 | 22 | Claude Code issues | | codex cli | 8,500 | 7 | Codex issues | | what is codex | 7,000 | 1 | Codex issues | | what is cursor | 4,300 | 22 | Cursor guides | | claude code usage | 3,600 | 30 | Token tracker | | agents.md | 1,400 | 53 | His AGENTS.md | | ai model comparison | 1,400 | 12 | OpenRouter page | | openrouter leaderboard | 350 | n/a | OpenRouter page | Searches a page per tool or a free tool could answer, and what Ben already has to build it from (Ahrefs, US). “what is codex” is the easiest of these: 7,000 searches a month at a difficulty of 1, with a small Swedish AI news site and a tools directory on page one. Ben’s “What makes a good AGENTS.md?” post is 198 words. A version with his own file, ready to copy, answers “agents.md” and its “template” and “example” searches in one page. ## AI search: the crawlers get every post, the answers go elsewhere ### What AI crawlers get Every AI crawler gets every post in full, about 4,300 words for “How I code with agents”. robots.txt doesn’t block any of the AI bots, and each post has NewsArticle structured data. | User agent | Homepage | A post | | --- | --- | --- | | Browser | 200, no post links in the HTML | 200, full text | | ClaudeBot | 200, no post links in the HTML | 200, full text | | Googlebot | 200, 34 post links and site structured data | 200, full text | | GPTBot | 200, no post links in the HTML | 200, full text | | OAI-SearchBot | 200, 34 post links and site structured data | 200, full text | | PerplexityBot | 200, no post links in the HTML | 200, full text | What www.bensbites.com serves for the homepage and "How I code with agents" to each user agent, 4 November 2026. Requests sent with each bot's user agent, not from the bots' own addresses. The homepage is different. Substack sends Googlebot and OAI-SearchBot, the crawler behind ChatGPT search, a full version with 34 post links. GPTBot, ClaudeBot and PerplexityBot get what a browser gets, with the post list loaded by JavaScript, so they find posts through the sitemap and the archive (24 links) instead. Ben can’t change either version. ### Who AI answers cite Across 30 answers from ChatGPT search, Google AI Mode and Google’s AI Overviews, Ben’s Bites is cited 4 times and mentioned twice. The Rundown is cited 5 times and mentioned 7 times. Every citation, for both, is the homepage. Ben’s Bites does best when the question says “builders” or “coding tools”, which is the audience the new homepage title names. Answers change from one run to the next, so ten questions are a sample. | Question | Ben's Bites | The Rundown | | --- | --- | --- | | best ai newsletter for non-technical people | Not named | Mentioned by ChatGPT and AI Mode | | best ai newsletters for builders and founders | Cited by ChatGPT and AI Mode | Cited by ChatGPT and AI Mode, mentioned by AI Overview | | best newsletter about ai coding tools | Cited by ChatGPT, mentioned by AI Overview | Not named | | best newsletter for learning to build with ai agents | Not named | Cited by AI Overview | | claude code vs codex for non-technical users | Not named | Not named | | how do i start coding with ai agents if i'm not technical | Not named | Not named | | how to use claude code if i can't code | Not named | Not named | | what are the best vibe coding tools for beginners | Not named | Not named | | what is the best ai newsletter | Not named | Cited by AI Mode, mentioned by the other two | | which ai newsletters should i subscribe to in 2026 | Cited by ChatGPT, mentioned by AI Overview | Cited by ChatGPT, mentioned by the other two | Whether each answer cites or mentions the newsletter, for 10 questions a Ben's Bites reader might ask (DataForSEO, US, 4 November 2026). AI Overview means Google's AI Overview. Not named also covers the 5 of 60 answers that didn't come back. On the four how-to questions, neither newsletter appears. The answers draw on Anthropic’s and OpenAI’s help pages, Zapier, Builder.io and Reddit. For “how to use claude code if i can’t code”, Google AI Mode cites a free course site called CC for Everyone, written for exactly the readers Ben writes for. The AI crawlers already get every word Ben writes, so the missing citations come from missing pages. The pages in New content are the fix: each one starts with a direct answer, compares tools by name and says when Ben last tested them. For the newsletter questions, the lists are the way in again. AI Mode’s sources for “what is the best ai newsletter” were The Rundown, Superhuman, TLDR, Medium, GitHub, Quora, Reddit and YouTube, and the GitHub list is already on the outreach list. ## Technical fixes, closest to revenue first The crawls found five problems. I’ve put them in order of how close each one sits to new subscribers, so the first one isn’t the one that affects the most pages. ### 1\. The catalog returns a 404 on every page When a page returns a 404, Google drops it from its results after a few recrawls, and the links pointing at it stop counting for anything. That’s happening to every catalog page at once, including the tutorials that brought in most of Ben’s Bites’ search traffic. Some of the work is already done. 95 of the catalog’s 201 articles were republished on the newsletter under the same address ending, such as “ChatGPT vs OpenAI API: Differences explained”. The old catalog addresses just don’t point at them. 1. Restore /tutorial/getting-started-accessing-chatgpt-and-setting-up-your-environments About 186 visits a month from searches for the ChatGPT interface. Rebuild it as a post and redirect it there. 2. Restore /tool/chatbot-arena Ranked second for "chatbot arena" (2,500 US searches a month). 3. 301 /blog/chatgpt-vs-openai-api-differences-explained redirects to www.bensbites.com/p/chatgpt-vs-openai-api-differences-explained One of 95 articles that already live on the newsletter. 4. 301 /role/executive redirects to www.bensbites.com/ The top result for "ben's bites newsletter". Those searchers want the homepage. 5. 410 /pricing The Pro membership it sold has closed. 6. 410 /tutorial/… (about 400 more) Tutorials with no traffic and no links. A 410 tells Google they're gone on purpose. Six catalog addresses and what I'd do with each. All six return a 404 today. **The fix:** one list of old and new addresses, applied as a Cloudflare bulk redirect, since the catalog’s hostname already runs through Cloudflare. Each old address gets a permanent (301) redirect to the page that replaces it, or a 410 when nothing does. I’d restore the 32 catalog pages that still had search traffic in Ahrefs’ November data as posts on bensbites.com, starting with the ChatGPT getting started tutorial and Chatbot Arena. Then crawl the old list again with `seo crawl --mode list`: every address should answer in one hop. ### 2\. The old domains send every address to the homepage Ben’s Bites has lived on other domains before. bensbites.co was the newsletter’s home, and news.bensbites.com was a community news site. Ahrefs counts 797 sites still linking to bensbites.co and 205 to news.bensbites.com. Both now redirect every address to the homepage, in two hops. Today bensbites.co/p/no-claude-for-claws 301 → bensbites.com/ 301 → www.bensbites.com/ (the homepage) news.bensbites.com/users/www7800betcom 301 → bensbites.com/ 301 → www.bensbites.com/ (the homepage) After the fix bensbites.co/p/no-claude-for-claws 301 → www.bensbites.com/p/no-claude-for-claws news.bensbites.com/users/www7800betcom 410 Gone Two old addresses that other sites link to, today and after the fix. [Google’s site move guide](https://developers.google.com/search/docs/crawling-indexing/site-move-with-url-changes) warns that redirecting many old addresses to the homepage “might be treated as a soft 404”, so the links to old posts may count for nothing. Three of the 77 bensbites.co post addresses that Ahrefs shows links to have an exact copy on the newsletter today, such as “No Claude for Claws”, and they still go to the homepage. news.bensbites.com has a different problem. 12 of its 15 most linked addresses are user profiles named after betting sites, each linked from about 230 sites (Ahrefs). They look like spam accounts with links built to them, and the redirect now points all of those links at Ben’s homepage. **The fix:** redirect each old address that has a real replacement straight to it in one hop, and return a 410 for the spam profiles (everything under news.bensbites.com’s users path). The 2023 issues with the most links, such as the ones on ChatGPT plugins and the wave of ChatGPT apps, are worth restoring if the old export still has them. Everything else can keep its homepage redirect. ### 3\. The cookbook is a separate site, and Google can’t read its sitemap The cookbook holds Ben’s best guide, and it’s set up worse than the newsletter. Its sitemap lists every page at localhost:8000, the address of the machine it was built on, so Google can’t use it. Its robots.txt and llms.txt return the homepage, and so does any made-up address. No page names a canonical. Its four glossary pages aren’t linked from its homepage. https://cookbook.bensbites.com/any-made-up-address 200 OK Browser tab: Ben's Bites Cookbook - Ben's Bites. Address: https://cookbook.bensbites.com/any-made-up-address. Status: 200 OK. I can’t code. I build with agents. The cookbook homepage, served for any address. An address that doesn't exist on the cookbook. It answers 200 OK with the homepage, so Google sees one more copy of the homepage instead of a missing page. Today cookbook.bensbites.com/cookbook/become-technical-in-1-day/ 200 OK, no canonical tag cookbook.bensbites.com/sitemap.xml http://localhost:8000/cookbook/become-technical-in-1-day/ cookbook.bensbites.com/any-made-up-address 200 OK, the homepage HTML After the fix cookbook.bensbites.com/cookbook/become-technical-in-1-day/ 301 → www.bensbites.com/p/become-technical-in-1-day www.bensbites.com/sitemap.xml lists the guide with every other post cookbook.bensbites.com/any-made-up-address 301 → www.bensbites.com/ The cookbook's main guide today, and after folding the cookbook into bensbites.com. **The fix:** I wouldn’t repair the cookbook as its own site. I’d republish its guides and glossary pages as posts on bensbites.com, starting with “Become Technical in 1 Day” and the questions in the table above. Then I’d redirect each cookbook page to its copy with a permanent (301) redirect, and send any other cookbook address to the homepage. “How I code with agents” already has a copy on the newsletter, so it only needs the redirect. After that, crawl the old cookbook addresses with `seo crawl --mode list`: each should answer with one 301 to a post. ### 4\. Most posts get no links from other posts A post with no links from other pages depends on the sitemap to be found, and gets none of the authority the rest of the site has built up. I checked every link in the body of all 270 posts in Ben’s Bites’ sitemap. October to December 2026 10 July to September 2026 33 April to June 2026 27 January to March 2026 29 October to December 2025 25 July to September 2025 28 April to June 2025 33 January to March 2025 52 October to December 2024 24 July to September 2024 9 - Linked from another post - No link from any other post Every post in the newsletter's sitemap, newest first. Lit tiles get at least one link from another post. Hover a tile for its title. 251 of the 270 posts get no link from any other post. The homepage and archive load their lists with JavaScript, and when they run, the homepage links to the newest 17 posts and the archive to the newest 24. Everything older is behind infinite scroll, and [Google doesn’t scroll](https://developers.google.com/search/docs/crawling-indexing/javascript/lazy-loading) to load more. For a daily news issue, that’s fine. It matters for the posts that answer a question someone will still search next year, such as “ChatGPT vs OpenAI API: Differences explained”, which ranks second for “chatgpt api vs chatgpt plus” without a single link from another post. **The fix:** Substack doesn’t let you change its templates, so the links have to go in the posts. I’d tag the evergreen posts (Substack already has a reviews tag page), write one “start here” post that links to the best of them, and link each new issue to one or two older posts on the same subject. Then rerun the check: the evergreen posts should all have at least one link. ### 5\. Old issues link to addresses that don’t exist The crawl found 26 addresses on [www.bensbites.com](http://www.bensbites.com/), linked from older issues, that return a 404. Some are Ben’s own posts from before the move to Substack that never came across. Others are posts on other newsletters, such as Ethan Mollick’s One Useful Thing, whose links now point at Ben’s own domain. They have the same address ending, so they look like links rewritten during the import. Today www.bensbites.com/p/prophecies-of-the-flood 302 → /prophecies-of-the-flood 404 Not Found Fixed www.oneusefulthing.org/p/prophecies-of-the-flood 200 OK (the post the issue meant to link to) A link in "BB Digest: ChatGPT has an alarm now", and the post it meant to reach. **The fix:** point each link to those 26 addresses at the original post, or remove it when the original has gone. It’s a small job, and it stops readers who click through from an old issue hitting a 404. ## Ship it all today, then let it settle None of this needs a sprint plan. Each fix is a separate job with a clear end, so coding agents can work on all of them at once: 1. Catalog redirects Export every catalog URL from Webflow or the Wayback Machine, on both catalog.bensbites.com and the old bensbites.com paths. Point the 95 articles at their newsletter copies, the name searches at the homepage, and return 410 for the rest. One Cloudflare bulk redirect list does it. 2. Restore the tutorials that ranked Rebuild the 32 catalog pages that still had search traffic as posts on bensbites.com, and redirect each old address to its new one. 3. Old domains Map bensbites.co and news.bensbites.com addresses to their exact new pages in one hop, and return 410 for the spam profiles. 4. Broken links in old issues Find every link in the archive that points at a bensbites.com address that 404s, and point it at the original source. 5. Fold in the cookbook Republish the cookbook guides and glossary pages as posts on bensbites.com, and redirect each cookbook page to its copy. Do the same for the sessions site. 6. Guides from X Split Become Technical in 1 Day into one post per question, starting with what a terminal and Bash are, and keep the full guide as the hub that links to them. 7. List outreach Draft a short, accurate pitch for each best AI newsletter list that ranks and leaves Ben's Bites out. 8. Comparison pages Turn the Vibe Coding Deathmatch into the best vibe coding tools hub and the five pair pages, each from Ben's own test, with his Factory role stated at the top. 9. Search titles on the posts that have them Retitle How I code with agents and What is vibe coding for the searches above, add a two-sentence answer at the top of each, and link every section to its own page. 10. Tool pages and free tools Write one page per coding agent from the archive, starting with Codex, and one post per free tool on Ben's personal site, plus his AGENTS.md as a template. Someone at Ben’s Bites still needs to pick which tutorials to restore and write the guides in Ben’s voice, because that’s the part readers subscribe for. The redirects and link fixes need only a quick check. Then I’d spend 30 days checking the results: 1. Day 0: Ask Google to recrawl Submit every redirected, restored and new URL to Index My Site, so Google sees the changes now instead of on its own schedule. 2. Day 3: Check the redirects Crawl the old catalog, cookbook, bensbites.co and news.bensbites.com addresses again. Each should answer in one hop with a 301 to a real page, or a 410. 3. Day 7: First impressions Look for the restored guides in Search Console, and impressions on "how to use claude cowork" and "claude code tutorial". 4. Day 14: Sign-ups from search Check Substack's subscriber sources for sign-ups from Google. 5. Day 30: New readers Compare free sign-ups from search with the month before. Those readers are what sponsors pay for. 6. Day 30: Ask the AI questions again Ask the same 10 questions for bensbites.com and therundown.ai with SEO Skill's AI answers research, and compare with today's 4 and 5 citations. ## Get the changes seen quickly The catalog pages are still in Google’s index, and Google takes its time to recheck a page that has been fine for years. I’d submit every redirected and restored URL to [Index My Site](https://indexmysite.app/) as it ships. That matters most for the catalog tutorials, where the next recrawl decides whether Google keeps the ranking or drops the page for good. ## 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. **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 domain we have used. Crawl each one with seo crawl, including old domains and subdomains. Check where each old address redirects. Crawl the old URLs the Wayback Machine has saved. Use seo_research to find the searches new readers make and who ranks for them, and the searches our competitors rank for that we don't. Ask AI assistants ten questions our readers ask with seo research ai_answers, for us and our biggest competitor. 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 your newsletter has moved platform or domain, crawl the old addresses too. In this teardown, that’s where the traffic went. If you’re on the Ben’s Bites team, I’m happy to go through any of this. And if you’d like the same teardown for your newsletter, [SEO Skill](https://seoskill.dev/) will run it for you.