Hey Guys,
Google just wrote down what it calls low-effort content. Four words: effort, originality, skill, accuracy. If you publish with AI, that matters this week.
That’s one of twelve plays in this edition. Here’s the spread:
Get customers fast. One operator lands a first $499/month client in 7 days with cold calls and a free-reviews offer. Another finds AI money in boring businesses by selling outcomes, not technology. Two paid-traffic case studies show buyer-intent Google and YouTube ads turning clicks into commissions.
Win attention. A reel hook framework built on four aligned elements. A YouTube recommendation method built on one word: relevant.
Rank and get cited. Google’s new quality rules, a 9,249-case-study meta analysis of what actually moved traffic, an exact-match domain play for local search, a bulk AI content playbook, and a cheap decision model for SEO grunt work.
One more, and I’ll be straight about it. The AI influencer model is in here because it’s everywhere right now, and the numbers are the creator’s own. Read it as a case study, not a promise.
A caution that applies to most of this edition: a lot of the results are self-reported, and several videos end in a pitch. I’ve flagged that where it matters. Treat every number as a test idea, change one thing at a time, and let your own data decide.
If you only have time for three, start with the Google low-effort rules, the 9,249-study test list, and the 7-day client plan. They’re the fastest to act on.
Let’s get into it.
1. Land Your First $499/Month Client in 7 Days With Zero Assets
SUMMARY
Balint Adorjan runs a $40K/month agency built on cold calls from a phone and laptop, with no ads and no Zoom calls. His argument is that beginners stall because they think they need a logo, a site, a CRM and case studies. They need a business owner with a problem and an offer that’s easy to say yes to.
His 7-day plan: pick one home-services niche and build a list of a few hundred businesses (day 1), learn one offer (day 2), practice one opener (day 3), call (days 4-5), follow up only on real conversations (day 6), then close (day 7). The offer is 25 free Google reviews from the prospect’s existing customers, in exchange for feedback. Owners already understand both “reviews” and “free.”
The mechanism is volume plus simplicity. He says 20 calls is a warm-up, not data, and he has had 150-call days that closed four $499/month clients. For first clients, proof you can start a conversation with a stranger and get paid matters more than squeezing the price.
KEY INSIGHTS
You need a business owner with a problem and an easy-yes offer. A logo, site, CRM and case studies are not required.
Which home-services niche you pick matters far less than picking one on day 1.
One offer beats ten: “25 Google reviews, free, for feedback.”
Don’t mention AI, automations or service bundles on the first call. Owners understand reviews and free.
The opener’s only job is to survive the first 30 seconds, using a specific gap like “you only have 14 reviews.”
He says 50 calls teach more than 20 more hours of YouTube.
20 calls is a warm-up. He has closed four $499/month clients in a 150-call day.
Follow up only with people who showed interest, and keep prospecting while you do.
ACTION CHECKLIST
Pick one home-services niche today and stop researching alternatives.
Build a list of 300+ businesses, with owner mobile numbers where possible.
Write your single offer: 25 free Google reviews from existing customers, for feedback.
Script and rehearse a 30-second opener built around the prospect’s low review count.
Make 50 calls without rewriting the script, then repeat the next day.
Follow up only with callbacks and partner conversations, and close your first client.
🎥 Watch the full video on YouTube: https://www.youtube.com/watch?v=ZiN-r8VToQ0
2. Build Reel Hooks That Stop the Scroll (and the Right Audience)
SUMMARY
Dominik’s thesis is that a reel hook is four aligned elements: the visual, the text on screen, the spoken line and the first six words of the caption. Together they have two jobs. They must give the right viewer instant clarity about who the reel is for and what it covers, and they must open a curiosity gap that keeps that viewer past the 2-3 second mark.
A high skip rate doesn’t always mean a bad hook. It can mean the algorithm showed the reel to the wrong people. The algorithm tests on viewers it thinks will like the reel, and if they stay it expands to similar people. When views stall at a few thousand, it often ran out of people who match. In his retention tests, clarity needs to land in the first half second.
For the curiosity gap, he warns that “nobody cares about you” yet. Hooks like “this insane thing happened to me” only work on followers, which he says describes about 70% of his clients. The fix is novel value, broken expectations or a challenge to the status quo, found by listing what everyone says about a topic and hunting for the angle nobody uses.
KEY INSIGHTS
A hook is four aligned elements: visual, text, spoken, and the first six caption words.
Every hook needs instant clarity (who it’s for) plus a curiosity gap (a question the viewer must answer).
A high skip rate can mean wrong-audience delivery, not a weak hook. His example is a 26M-view creatine reel that a true-crime viewer would skip.
In his tests, put the text hook on screen from frame one, with no animation delay.
If most of your reach is followers, context-dependent hooks are the likely cause. He says it affects about 70% of his clients.
Reused hook-list formulas get swiped faster, especially when a viewer sees the same hook twice in a session.
The reliable curiosity gap comes from novel value, broken expectations or a challenged status quo.
One topic with different angles can yield 10-30 distinct reels.
ACTION CHECKLIST
List every angle your niche already uses on your next reel topic.
Find one novel or contrarian angle almost nobody covers.
Write the single question you want the viewer asking after the hook.
Put the text hook on screen in frame one.
Align the visual, text, spoken line and first six caption words to the same viewer.
Check that your profile delivers what the reel promised.
🎥 Watch the full video on YouTube: https://www.youtube.com/watch?v=3e9y95qsKlg
3. Use an Exact-Match Domain to Win Local Search and Cut Ad Costs
SUMMARY
David Quaid, on Edward Sturm’s podcast, makes the case that a keyword-rich, domain-style business name can lift both Google Business Profile and organic rankings. A domain like escapewebdesignnyc.com can serve as the profile name, backed by a registered trading name (a DBA in the US). He says Google usually accepts it. He treats the domain as a ranking multiplier rather than a signal: once people search your brand plus the keyword, the keyword becomes part of your topical authority.
He lays out five profile levers: name and location, services matched to search phrases, matching web pages, linking those pages to the profile, and links. He says a mapped service page can start ranking within hours. He also predicts an EMD page aimed at your costliest paid keyword could lift quality score from about 3 to 6-7 and cut paid costs 20-40%, possibly up to 60%. That figure is his speculation, and a colleague’s team plans to test it.
He’s candid on limits. The 2024 API leak shows EMDs were demoted, “near me” and “best” aren’t read from a domain, and established brands are hard to displace. His structural advice is to use subfolders over subdomains, avoid duplicating about and contact pages across brand sites, and expect Google’s video verification to be the hardest step.
KEY INSIGHTS
An LLC name, a trading name (DBA) and a trademark are three different things. The trading name goes on invoices and footers.
Quaid says Google usually accepts a domain-style name as the Business Profile name.
Google now requires video proof of a real, signed premises, which blocks many remote agencies.
His five profile levers: name and location, services matched to search phrases, matching pages, linking pages to the profile, and links.
A page mapped to your profile can start ranking within hours, he says.
Point the EMD page at your most expensive paid keyword. His untested estimate is a 20-40% cost saving.
“Near me” and “best” in a domain aren’t read as text, so they don’t add relevance.
Use subfolders, not subdomains, because subdomains share authority poorly.
ACTION CHECKLIST
Confirm your trading name is registered and matches your brand and domain.
Identify your costliest paid-search keyword and check whether a dedicated page exists for it.
Build that page and link it from your main site, without cloning your about and contact pages.
Map each service page to its matching Google Business Profile.
Put your map and profile on the contact page.
Test the EMD idea on one keyword and track paid cost per click before scaling.
🎥 Watch the full video on YouTube: https://www.youtube.com/watch?v=Fo-aGXG3umM
4. Find AI Money in “Boring” Businesses (and Validate It in 7 Days)
SUMMARY
Jeff Herschy argues most people chasing AI agents and SaaS are looking in the wrong place. The opportunity sits in businesses still running on phone calls, spreadsheets and paperwork. His thesis: AI made building cheap, so the problem is the opportunity, and the hard question is what’s worth building.
His filter has four parts: pain, frequency, value and AI leverage. Pain is how frustrating the problem is, frequency is how often it happens (50 times a week beats once a month), value is what it costs in leads, revenue or hours, and leverage is whether AI can do the repetitive part while a human handles exceptions. The rule is to sell the outcome, not the technology: “I help contractors follow up with every new lead automatically,” not “I build AI agents.”
He validates before building. Talk to three owners, ask them to show how they handle the problem today, how often, who owns it and what a miss costs. The first version can be manual. He prices on the value of the problem, not hours: 10 clients at $1,000/month or 20 at $500 is $10K/month.
KEY INSIGHTS
The problem is the opportunity. AI just made building it cheap.
Using ChatGPT for emails is different from building AI into a workflow.
His four-part filter is pain, frequency, value and AI leverage.
Sell outcomes, not technology: “follow up with every lead automatically.”
In professional services, AI handles the repetitive work around decisions, never the decisions.
Don’t ask “would you use an AI tool?” Ask owners to show how they handle it today and what a miss costs.
Your first version can be manual. Use AI’s low cost to test faster, not just build faster.
Price on the problem’s value: 10 clients at $1,000/month or 20 at $500 equals $10K/month.
ACTION CHECKLIST
Pick one boring industry you already understand.
Talk to three owners and ask them to walk through the workflow without pitching.
Ask how often it happens, who owns it, what a miss costs and what it costs per month.
Score the problem on pain, frequency, value and AI leverage.
Build the simplest possible version (manual if needed) and show it to the same owners within a week.
Rewrite your pitch as an outcome with no mention of agents or AI.
🎥 Watch the full video on YouTube: https://www.youtube.com/watch?v=HWjSU0aIKfM
5. Google’s New “Low-Effort” Rules: What to Fix Before the Next Update
SUMMARY
Matt Kenyon reports that on October 1 Google updated its helpful content guide and its AI content guidance. The guide now names “main content” (text, images, video, tools, reviews, headings) and lists four rater criteria: effort, originality, talent or skill, and accuracy. Fake authors, AI avatars and invented credentials are called deception, and fact-checking AI output, including titles, meta descriptions and alt text, is “critical.”
Raters don’t rank pages directly, but they train the systems that do, so the criteria show what those systems learn to reward. Kenyon cites the pattern that guideline changes have preceded major updates (Panda, Penguin, Medic). Google has run four spam updates this year versus one in all of 2025, and Google’s Gary Illyes said scaled content now worries them more than link spam. His inference, which he labels as opinion, is that an AI-focused update is coming.
His examples of passing: Crutchfield’s own photos, diagrams and videos, and Rilla’s study of 10,000 recorded sales appointments. His warning list: scaled templated pages, fake authors, self-promotional “best X” listicles and paid list placements. His advice is to start with something AI doesn’t have, verify every claim yourself, read every word, use a real author, disclose AI use and not publish beyond your expertise.
KEY INSIGHTS
The four criteria are effort, originality, talent or skill, and accuracy.
Google calls fake authors, AI avatars and fake credentials deception.
Fact-checking and reviewing AI output, including titles, meta descriptions and alt text, is now “critical.”
Raters train the ranking systems, so their criteria show what gets rewarded.
Four spam updates this year versus one in all of 2025, with shrinking gaps.
Citing sources doesn’t replace original effort.
In his 116-listicle CRM study, the writing brand ranked first only about a quarter of the time, and ChatGPT recommended only competitors in over a third.
Original data (Rilla’s 10,000 recorded appointments) is the kind of asset an LLM can’t copy.
ACTION CHECKLIST
Start each piece with something AI lacks: your data, a test, a client call or a real opinion.
Verify every number, quote and link against the original source.
Read every word of AI output and rewrite hedging, machine-sounding passages.
Put a real author with a clickable bio on every article.
Add a line on how you used AI.
Audit templated pages that differ by a few words, and add original photos, diagrams or tools to your best pages.
🎥 Watch the full video on YouTube: https://www.youtube.com/watch?v=qMUAU7E3dCA
6. The 5-Part AI Playbook Behind a Site’s Jump From 1,700 to 22,000 Monthly Visitors
SUMMARY
Avi reports growing a site from about 1,700 to 22,000 monthly active visitors in a year, mostly US traffic. His most effective move was using AI to publish articles in bulk, over 838 in about 18 months. His view is that volume built topical authority, but the articles weren’t thin: each was in-depth, many had custom images or HTML elements, and all were fact-checked with Perplexity. He expects only about 10% of 800 articles to bring traffic and says you can’t predict which.
The other four tactics compound the first. Custom images drove about 12,000 image-search clicks in a month. Winning posts became videos through NotebookLM’s video overview, with some reaching 100K, 59K and 56K views. AI-assisted outreach produced links from Verywell Health (DR 90) and Melanoma Canada. AI also flagged technical and on-page fixes, which led him to build RankSite, a fast-loading HTML site builder.
These are his own figures, and he promotes his own tools (WordRocket and RankSite) along the way, so treat the numbers as a case study and not a benchmark.
KEY INSIGHTS
He reports 1,700 to 22,000 monthly active visitors in a year, mostly from the US.
He published 838+ AI-assisted articles in about 18 months to build topical authority.
Quality guardrails: in-depth articles, custom images or HTML, and Perplexity fact-checking on each.
He expects roughly 10% of articles to drive traffic and says you can’t predict which.
Custom images earned about 12,000 image-search clicks in one month.
Top-performing posts were turned into videos via NotebookLM, with some at 100K, 59K and 56K views.
An AI skill that shortlists backlink opportunities earned links from Verywell Health (DR 90).
Some figures in the source are inconsistent (the title claims 700K users), so verify before quoting.
ACTION CHECKLIST
Pick a niche where you can publish dozens of in-depth articles, not thin ones.
Add a fact-check step and custom images to every article.
Track which ~10% of posts earn traffic and build more around them.
Paste winning articles into NotebookLM, generate video overviews and link back to the post.
Use AI to shortlist backlink targets, then send personalized outreach.
Run an AI technical and on-page audit on your site.
🎥 Watch the full video on YouTube: https://www.youtube.com/watch?v=0KIivqLvFc8
7. 9,249 SEO Case Studies: The Tactics Worth Testing (and the Ones to Skip)
SUMMARY
Edward Sturm reacts to a Reddit meta study that screened 30,000+ documents down to 9,249, extracted results with two models, dropped any number not literally in the source, and graded each tactic on evidence and actionability. Sturm agrees with some of it and disagrees with some, and frames it all as ideas to test.
Classic SEO showed the strongest numbers: all-caps titles (+17.5% in two tests), removing synonyms from titles (-27%), price in the title (-15%), digital-PR editorial links (+71%), and orphaned posts gaining up to 64 positions after internal links. Disavow helped (+80%) only with a manual penalty and showed no measurable effect otherwise. AEO/GEO evidence was the thinnest: statistics (+37% citations), named-source quotes (+22-41%), a direct answer first (+280%, one practitioner), author names (+113%) and human editing of AI text (+67%).
The key caveat: nearly every GEO number comes from bundled changes, and no controlled llms.txt test appeared in the 9,249 documents. Sturm couldn’t isolate any single tactic backed by more than one controlled test, so test one change at a time.
KEY INSIGHTS
Two independent tests found all-caps title tags lifted organic traffic 17.5%.
Removing synonym keywords to shorten titles cut traffic 27%, the biggest negative from a single test.
Price in the title tag cut traffic 15%. Sturm says bottom-funnel pages often do better when people click to learn the price.
Digital-PR editorial links were associated with a 71% organic traffic lift.
Disavow helped (+80%) only with a manual penalty. Otherwise there was no measurable effect.
Internal links to orphaned posts gained up to 64 positions.
AI citation boosts: statistics +37%, named-source quotes +22-41%, author names +113%, human editing +67%.
AI text on specific pages raised citations 73%, while sitewide AI text cut sessions 35%.
ACTION CHECKLIST
Test one title-tag change at a time, and remove price from titles on a test page.
Find orphan pages and add internal links to them.
Open key pages with a direct answer before the evidence.
Add statistics and named-source quotes with the source placed next to each claim.
Put a real author name on posts and have a human edit every AI draft.
Skip disavow unless you have a manual penalty, and treat GEO figures as test ideas, not rules.
🎥 Watch the full video on YouTube: https://www.youtube.com/watch?v=q3zknvoFuNc
8. Inside the AI Influencer Affiliate Model: Realism, Volume and the Content Puzzle
SUMMARY
Pepe says AI influencer pages generated over $6M in revenue for his brand in 10 months, and he lays out the model step by step. Those numbers are self-reported, and the video ends in a pitch for his Discord coaching. His claim is that startup cost is about $100-200/month in software, with no inventory and no face on camera. He recommends health or beauty because they convert best and have the largest audience, with a product that already has many reviews.
His process starts with realism. Source a Pinterest reference of a niche-authority look, generate a character sheet with four angles in GPT Image 2, add a single-light-source room reference, then generate video in Seedance with a Claude skill that encodes film-style direction. For ideas he uses a “content puzzle” of hook, body, language, demographic and pain point: take a viral video, keep the structure, and swap the pain point or demographic so new people see it.
For monetization he describes an Amazon Associates bio link, growth content first (recipe videos with a “comment recipe” CTA and automated DMs), then sales content once a video pops. To scale he hires Filipino video editors and keeps scriptwriting in-house, claiming 80-100 posts a day. Verify platform rules, affiliate terms and health-claim compliance before copying any of it.
KEY INSIGHTS
He claims $6M+ revenue and $1M+ profit in 10 months. These are his own figures.
Startup cost cited: roughly $100-200/month in software, no inventory.
Health and beauty convert best and have the biggest addressable audiences, he says.
Realism is the foundation: a four-angle character sheet fixes plastic skin and inconsistent faces.
A reference photo with one light source makes video lighting look real.
His “content puzzle” splits a video into hook, body, language, demographic and pain point. Swap one or two pieces to re-run a winner.
He trains a Claude skill on Reddit, Facebook-group and Amazon-review language to rewrite scripts to new pain points.
Funnel order: growth content first, sales content after a video pops.
ACTION CHECKLIST
Choose a niche and a product that already has many reviews.
Build a character sheet and a single-light-source room reference for your avatar.
Generate a test video and check skin texture, lighting and face consistency.
Break one recent viral video into hook, body, language, demographic and pain point.
Rewrite the script for a new pain point using real customer language from reviews and forums.
Check platform and affiliate rules, and health-claim rules, before launching.
🎥 Watch the full video on YouTube: https://www.youtube.com/watch?v=x0p9bWvlZvs
9. Get Recommended on YouTube by Making Every Video “Relevant to Me”
SUMMARY
Brian G. Johnson says the fastest route out of a stalled channel is getting recommended, which he calls the “wave strategy.” His example is a video at 17,000 views on day three despite unremarkable retention. The key word is “relevant,” specifically “relevant to me” from the viewer’s side. His old videos flopped because they were too narrow, so large parts of the audience disqualified themselves before clicking.
The mechanism is that the algorithm tests a new video on viewers it thinks will like it, and the title, topic, thumbnail and opening one to two minutes decide how many click and stay. A title about running ten miles a day excludes most people, and he estimates a title about presenting on camera loses about 40% of viewers. His fix is to keep making similar videos but change the angle and the entry point, opening broad and narrowing to the specific point later. He uses YouTube’s Ask Studio to test whether a topic is broad.
On metrics, he argues the key signal is watch time per impression, which combines click-through rate and average view duration. His broad video held a stable CTR for three days, and a longer runtime lifted average duration despite 26% retention. Two of his last 10 videos now outperform strongly, one reaching 91,000 views on a sub-1,000-subscriber channel.
KEY INSIGHTS
Relevance (“is this for me?”) decides whether viewers click.
Narrow titles and topics make large parts of the audience disqualify themselves.
He estimates a title about presenting on camera loses about 40% of potential viewers.
You can keep making similar videos by changing the angle or entry point.
Design the first one to two minutes for a broad audience, then narrow to the specific point.
He calls watch time per impression the main satisfaction signal. It combines CTR and average view duration.
A stable CTR across three days, instead of the usual decay, was his encouraging sign.
Retention was 26%, but a longer video (about 20 minutes) lifted average duration.
ACTION CHECKLIST
Rewrite your next title so most people in your niche feel it applies to them.
Ask whether the title excludes any large group, and fix it.
Open the first one to two minutes broadly before narrowing to the specific point.
Describe your channel and planned video to Ask Studio and ask why viewers might not click.
Track CTR, average view duration and watch time per impression on each upload.
Compare your last 10 videos and study the two or three that outperform.
🎥 Watch the full video on YouTube: https://www.youtube.com/watch?v=8axfDPDlcxg
10. A ClickBank Case Study: $1,429 in One Day From Google and YouTube Ads
SUMMARY
Simon says a brand-new Google Ads account earned about $1,429 in affiliate commissions on October 6, with roughly $3,300 in commissions in the first week of a campaign started October 1. The spend was $671 for 665 clicks (about $1 each) and 3,400 video views, split between YouTube video ads and Google search ads. The offer was a ClickBank woodworking plans product paying about $60 per sale. These are his own figures, and a good share of the video pitches his done-for-you traffic service.
The structure is simple: an opt-in page offering free woodworking plans (50 free plans) for an email address, ads driving to that page, a list, and at least a 30-day email follow-up sequence. Targeting was the US, Canada, UK and Australia, ages 30-65, via keywords and interest audiences. The mechanism he emphasizes is buyer intent: search ads reach people already looking for a product or solution.
He normally recommends offers paying $300-600 per sale and tested a lower-payout product here. He also mentions Google Display Network banners and says he tests multiple offers to see which perform best.
KEY INSIGHTS
He reports $1,429 in commissions in a single day, from a campaign started five days earlier on a new account.
Reported spend was $671 for 665 clicks (about $1 each) and 3,400 video views.
The campaign mixed YouTube video ads and Google search ads.
The offer paid about $60 per sale through a video sales page.
The structure is opt-in page, ad traffic, email list and a 30-day follow-up series.
Lead magnet: free woodworking plans in exchange for an email address.
Targeting: US, Canada, UK and Australia, ages 30-65.
He normally recommends $300-600 offers, and results here are self-reported.
ACTION CHECKLIST
Choose an affiliate offer and confirm its payout per sale (favor several hundred dollars).
Build an opt-in page with a free lead magnet tied to the offer.
Write a 30-day email follow-up sequence.
Launch search and YouTube ads on buyer-intent keywords with a small test budget.
Track clicks, cost per click and commissions daily.
Verify any traffic-service claims before paying for anything.
🎥 Watch the full video on YouTube: https://www.youtube.com/watch?v=FDVn95dQXhc
11. Why Your Google Ads Get Clicks but No Sales: The Message-to-Market Fix
SUMMARY
The creator reports 34 sales and over $2,000 in three days promoting a new digital affiliate product with Google paid-click ads and no video. These are self-reported results, and the video ends in a WhatsApp and email pitch for his coaching and traffic service. His premise is that older Google Ads affiliate tactics failed because of weak intent. People were curiosity clickers who never bought. The fix is to send buyer-intent clicks, from people already searching for a solution, to a funnel.
He lists the requirements as a laptop, Wi-Fi and 20-30 minutes a day, and the model is faceless with no product creation or coding. His income math: 24 sales a week at $100, 12 at $200 or 5 at $500 reaches about $10K/month. His central diagnosis is that 90% of people match the wrong ad to the wrong audience, so the real work is message-to-market match: audience gender, language patterns and copy on the landing page or quiz funnel.
The process he describes is to install one affiliate campaign (e.g. ClickBank), launch the Google campaign and scale toward $350-500/day once it shows profit. He also cites a $50 spend yielding one sale and $250 profit within 24 hours with upsells and downsells.
KEY INSIGHTS
Buyer-intent search clicks outperform curiosity clicks.
He claims 34 sales and $2,000+ in three days on a new digital product with no video ads.
Income math: 24 sales/week at $100, 12 at $200 or 5 at $500 each reaches about $10K/month.
One $200 sale a week adds about $800/month. Five a week adds about $4,000.
He says 90% of people match the wrong ad to the wrong audience.
Message-to-market match depends on audience gender, language patterns and landing-page or quiz-funnel copy.
A reported $50 spend returned one sale and $250 profit in 24 hours with upsells and downsells.
Results shown are testimonials and self-reported, so test small before scaling.
ACTION CHECKLIST
Write down exactly who your affiliate offer is for and what problem they search.
Define the audience’s gender, language and pain point before writing any ad.
Target searches where people are already looking for a solution.
Build a landing page or quiz funnel in that audience’s own language.
Run the income math at your payout and a realistic conversion rate.
Track daily spend, sales and profit, and scale only after one campaign is profitable.
🎥 Watch the full video on YouTube: https://www.youtube.com/watch?v=PpJyaMBTz6k
12. Cut SEO Decision Costs With Jev: A Cheaper Brain Inside Your Claude Workflow
SUMMARY
Nico introduces Jev, a decision model from Type Safe, founded by Diogo Almeida, a former OpenAI researcher. It doesn’t write text or code. It answers fast: pick from a list, give a 0-to-1 score, or say yes/no. That fits SEO, which he describes as many tiny decisions: should this page link to that one, do these keywords share intent, is this site worth a backlink. Spending expensive LLM tokens on each adds up.
It has limits: weak on math and dates, easily overwhelmed by large data, and wrong sometimes even though it can’t fabricate. You can reach it through type-safe.ai (he says $5 free credit, though he had sign-in problems) or OpenRouter, his preference, where a few dollars goes far. Claude or Codex with OpenRouter access can call it on request.
He shows three uses. Internal linking on a 135-page site returned confidence-scored link suggestions with reasoning, a dashboard and a CSV, and you can auto-apply anything above 0.8 confidence. Cannibalization checks use Search Console exports (3+ months) to flag which page should win a query. ChatGPT visibility reported mentions in 11 of 40 answers and recommendations in two. His view is that Jev covers one of six steps in a workflow and is useful mainly if you run many sites and burn credits. Spot-check its output.
KEY INSIGHTS
Jev is a decision model, not a chat model. It can’t write text or code.
It answers by choosing from a list, scoring 0 to 1, or giving yes/no.
It’s weak at math and dates and loses accuracy on big data, so split work into small questions.
It can’t invent facts, but it can be wrong, so human review still matters.
Access it via type-safe.ai (about $5 free credit) or OpenRouter, where a few dollars goes a long way.
Internal-link run: about 135 pages read, with confidence scores and an auto-apply rule above 0.8.
Cannibalization checks need 3+ months of Search Console data, and the flags still need human review.
It covers one of six steps in an internal-linking workflow, so it’s a cost-saver, not a replacement for Claude or ChatGPT.
ACTION CHECKLIST
Create an OpenRouter account and add a few dollars of credit.
Confirm your site doesn’t block AI crawlers before running the analysis.
Run the internal-link analysis and export the CSV.
Spot-check high-confidence suggestions, including anchor text and the target page.
Set a confidence threshold (e.g. 0.8) before auto-applying any links.
Export 3+ months of Search Console data, run the cannibalization check and review each flagged conflict.
🎥 Watch the full video on YouTube: https://www.youtube.com/watch?v=3cX55E5ILyU
Talk soon,
Tony