Get found on Google. Get quoted by ChatGPT. Every week, on autopilot.

Paperboy reads your own Google Search Console data, finds the searches you are one good article away from winning, and writes that article: researched from real sources with Claude's reasoning models, every claim fact-checked, published to your site. Then it watches what ranks and does it again.

  • Learns from your Google data
  • Every claim checked against its source
  • Written to be quoted by AI answers

No credit card. You see the finished article before we ask about a plan.

  1. 1
    Paste your address
    We read your site and describe your business in plain words. You fix anything wrong.
  2. 2
    Connect Search Console
    Optional. We see which searches already find you.
  3. 3
    Pick how articles go live
    Hosted blog, WordPress, Ghost, an AI site builder, or a webhook.
  4. 4
    Get your first article
    Researched, written, fact-checked and published in under ten minutes.

An SEO feedback loop that runs every week

Most AI writers stop at the draft. Paperboy keeps going.

1. Reads your Google data

Every day it pulls your Search Console numbers and finds the searches where you sit on page two, the questions you have no page for, and the topics next to what already gets clicks. It tells you why it picked each one.

2. Researches like an analyst

Claude's reasoning models run up to six web searches, open the best sources and save their exact text. Your own pricing and product pages count too, so it never guesses what you sell.

3. Fact-checks every claim

Every number must appear in the source it links to, and every cited sentence is compared with that source's own words. One rewrite if anything fails. Still failing means it waits for you, not your readers.

4. Learns from what ranks

It watches each article's clicks and position, refreshes the ones stuck on page two, and uses what worked to pick the next topic. The longer it runs, the sharper it gets.

Built for AI answers, not just blue links (GEO). ChatGPT, Perplexity, Claude and Google's AI Overviews quote pages that answer a question plainly and back it up. Every Paperboy article puts the answer first, links each fact to its source, says when a source sells something, and ships with FAQ and Article structured data. Check if AI tools can read your site.

A real article it wrote

For Riot Digital (riotdigitalai.com), a test customer. Quality 82 of 100 · 7 sources · made in 3 minutes for $0.31 of AI time

AI Email Assistant Cost Comparison for Contractors

A $297 per month managed AI email assistant pays for itself if it gives a contractor back about four hours a month: $297 divided by a $75 hourly value is 3.96 hours, using the $75 hourly rate in alfred_'s ROI calculator. Software tools cost far less, such as Superhuman at $30 a month, but you set them up and run them yourself, and a human executive assistant costs $60,000 to $120,000 a year fully loaded. Both of those pages come from alfred_, which sells its own AI email assistant, so treat the figures as a seller's numbers.

Quick version

  • The $297 per month figure is the price this article tests, with a setup fee on top. Riot Digital publishes this site, so I am an interested party. Confirm current numbers on the Riot Digital pricing page.
  • Before setup, $297 a month breaks even at about 0.9 hours of work recovered per week at a $75 hourly value. The math is below.
  • Cheaper software runs from about $7 to $36 a month for sorting and suggested replies. It does less of the work for you.
  • A custom build or a human hire costs far more up front or per year than the managed plan.
  • The premium is worth it only if you will use the hours you get back. In my view, that is the real test.

What each option costs

This table repeats published prices from sellers. Annual figures are my arithmetic, shown in each cell.

OptionPublished priceAnnual cost
Majordomo (sells its own software)$9 per month flat12 times $9 is $108
SaneBox$7 to $36 per month12 times $7 is $84, up to 12 times $36 is $432
Superhuman$30 per month12 times $30 is $360
Microsoft 365 Copilot$30 per user per month on top of the M365 subscription12 times $30 is $360, plus M365
Managed AI employee (the $297 plan tested here)$297 per month plus setup, confirm on the pricing page12 times $297 is $3,564, plus setup
Custom-built assistant (Appwrk sells development)$8,000 to $12,000 for an MVP, $30,000 to $48,000 enterprise-gradeOne-time build cost
Human executive assistant$60,000 to $120,000 per year fully loadedSame as the published range

One more reference point: Majordomo's own guide says most AI email tools cost $25 to $50 a month. Majordomo sells a $9 tool, so it has a reason to call that range expensive.

What the cheaper tools do and do not do

Most low-cost tools work inside your inbox. They sort, filter, and suggest. For example, Gmail's Smart Reply suggests three short responses. That helps with quick answers. It does not chase a client who never replied, and it does not book the estimate.

A managed service is a different kind of purchase. Riot Digital describes its product as an AI employee that handles email, calendar, and follow-ups. The AI Employees page lists what that covers. Riot Digital sets it up for you, which is part of what the extra money pays for.

My take: if your only problem is a messy inbox, a $7 to $30 tool is probably enough. If your problem is missed follow-ups and appointment back-and-forth, software that only suggests replies leaves most of that work on your desk. This is my judgment, not a sourced fact.

How many hours can you get back

The sourced numbers are about office workers in general, not contractors. Treat them as a rough starting point.

My take: many contractors are on job sites, not in an inbox all day, so the 50 messages per day case may not describe you. Count your own messages for a week before you trust any of these percentages. That is an opinion, not a finding.

Break-even math

Start with the formula. Halo Technology Lab, a UK firm that sells automation consulting, gives it as Annual ROI = ((Annual Savings minus Annual Cost) divided by Annual Cost) times 100.

Next, pick an hourly value. The $75 rate is from alfred_'s calculator. Your own rate may be higher or lower. The same page says to divide your gross annual salary by 2,080 to get a base hourly figure. An owner who does not pay themselves a salary can still estimate what an hour of billable work earns.

Here is a made-up example, not real data. It uses the $75 rate, the 3.3 hours per week from the table above, and the $297 monthly price. It leaves out setup because I have no sourced setup figure.

Hours saved per weekYearly value at $75 per hourYearly costROI
3.33.3 times $75 times 52 is $12,87012 times $297 is $3,564($12,870 minus $3,564) divided by $3,564 is 2.61, so about 261%
1.65 (half of that)1.65 times $75 times 52 is $6,43512 times $297 is $3,564($6,435 minus $3,564) divided by $3,564 is 0.81, so about 81%

The break-even point is $3,564 divided by $75 times 52, which is $3,564 divided by $3,900, or about 0.9 hours a week. Add your setup fee to the cost side and the break-even moves up a little. Halo says well-chosen automations typically deliver 200% to 500% ROI, so the first row sits inside that range and the second falls below it.

Compare that with a $30 tool. At a $75 hourly value, $30 divided by $75 is 0.4 hours, or 24 minutes a month. It is much easier to clear that bar. The question is how much work it takes off your plate once you clear it.

A caution about the big numbers you will see elsewhere. alfred_'s calculator claims roughly 75x return on a $25 per month tool and over $90,000 a year in savings at $300 per hour. Those are marketing figures from a seller, and the page itself is built to promote the purchase. I would not plan around them.

If you want a floor before you buy anything, Halo suggests a process should consume at least 5 hours per month to justify automating it. Email and follow-ups usually clear that for a small contractor, but check yours.

Versus hiring a person

The human option is the most expensive one in the table. A human executive assistant costs $60,000 to $120,000 per year fully loaded. At the low end, $60,000 divided by the $3,564 yearly managed cost is about 16.8, so the person costs nearly 17 times as much.

That comparison is not fair in every case. A person can show up at a job site, call a supplier, and handle problems nobody planned for. alfred_ claims an AI assistant handles 80% to 90% of the same email work as a human assistant, and I treat that as a vendor claim. The honest comparison is email, calendar, and follow-ups only, and a full-time hire is probably more than most small contractors need.

Versus building your own

A custom build is a real option, but it is mostly a poor fit for a small contractor. Appwrk, which sells AI email assistant development, puts MVP builds at $8,000 to $12,000 and enterprise-grade builds at $30,000 to $48,000. At the low end, $8,000 divided by $3,564 is about 2.2, so a basic build costs over two years of the managed plan. You would also still need to maintain it. Appwrk says a typical SaaS build takes 6 to 10 weeks, which is a long time to wait.

When the premium is and is not worth it

In my view, the $297 plan makes sense when:

  1. You lose leads or jobs because replies and follow-ups go out late.
  2. Calendar back-and-forth takes real time each week.
  3. You will spend the hours you get back on billable work or on time off.
  4. You would rather pay someone to set it up than spend a weekend on it.

It probably does not make sense when your inbox is light, when you only need sorting, or when you enjoy tinkering and would happily run a $9 tool. This is my judgment, not a sourced fact.

A simple way to test it

  • Track your email, scheduling, and follow-up time for one week.
  • Multiply those hours by 52, then by your own hourly value.
  • Compare the result with 12 times the monthly price, plus setup.
  • Ask any vendor what happens if the hours do not materialize.

If the numbers hold up with your own figures, a managed plan is easier to defend. If you want to run your numbers against Riot Digital's setup fee and monthly price, the pricing page is the place to start, and you can Book a call if you want to talk through whether it fits your business.

Frequently asked questions

Is a $297 per month AI email assistant worth it for a small contractor?

It can be. At a $75 hourly value, $297 divided by $75 is 3.96, so it breaks even at about four recovered hours a month before setup. Whether you actually recover those hours depends on your inbox and how much follow-up work you do.

How much do cheaper AI email tools cost?

Published prices run from $9 a month for Majordomo to $7 to $36 for SaneBox and $30 for Superhuman. Majordomo, alfred_ and the other sellers behind those pages have reasons to favor their own prices, so check each vendor directly.

How much does a custom-built AI email assistant cost?

Appwrk, a company that sells AI email assistant development, says MVP builds typically cost $8,000 to $12,000 and enterprise-grade builds $30,000 to $48,000. That is a one-time build cost, and it does not include maintenance.

How do I calculate ROI on an AI email assistant?

Use Annual ROI = ((Annual Savings minus Annual Cost) divided by Annual Cost) times 100. Estimate savings as hours saved times your hourly value times 52, and count setup as part of cost.

Its fact check

Passed the fact check on the first try. 9 small notes, none blocking.

Every claim has a sourcePass
Numbers match their sourcesPass
Sources say what the article saysPass
Sounds like the customerPass
No banned claimsPass
House stylePass

Its sources

See it live

What your dashboard looks like a few months in

An example Paperboy dashboard with sample data: clicks and impressions rising over 28 days, why each topic was picked, the articles coming next and how recent ones rank

An example account with sample data (a fictional heating company, a few months in): traffic from Search Console rising, the reason behind every topic, your own ideas queued first, and how each article ranks. Your dashboard fills in the same way with your real numbers.

How Paperboy compares

From each company's own pricing page, October 4, 2026
ToolStarting priceWrites and publishes on its ownPicks topics from your Search ConsoleChecks each claim against its sourceHolds articles that fail
Paperboy$49 a month per site for 3 articles a weekYesYes, picks every topic from itYes, every number and citation, every planYes, held for you
Outrank$99 a month for 30 articlesYesNot listedHuman fact-check add-on, $1,399 a monthNot listed
Koala AIFrom $25 a month (word-based)Yes, autopilot from $99 a monthYes, from $49 a monthCitations when you want them, from $49 a monthNot listed
ArvowFrom $39 a month, billed yearly (limited-time price)YesUsed to submit articles for indexingHuman editors in a service from $2,000 a monthNot listed
SurferFrom $49 a month, billed yearlyNot listed (writing and optimization editor)Not listedNot listedNot listed
JasperFrom $59 a month, billed yearlyNot listed (marketing writing platform)Not listedNot listedNot listed

"Not listed" means the company's pricing or home page does not mention it; the tool may still do it, so check before you buy. Sources: Outrank, Koala AI, Arvow, Surfer, Jasper. Full comparison.

Can ChatGPT and Perplexity even read your site?

One line in a robots.txt file can block AI search tools without anyone noticing. Our free checker reads your robots.txt and nine other basics in about ten seconds and tells you exactly what to fix. No sign-up.

Pricing

Per site. Every extra site is 20 percent off. Yearly billing gives two months free.

Starter

$49 a month
or $490 a year

3 articles a week

About $3.77 an article

Three articles a week. Right for a new site building its base.

Pro

$179 a month
or $1,790 a year

14 articles a week

About $2.95 an article

Two articles a day. Right for covering a whole market quickly.

Every plan includes the fact check, all publishing methods, Search Console learning and the weekly report. Start with one free article.

Questions

What does it actually do each week?

Paperboy picks a topic from your Google Search Console data (searches where you sit on page two, questions you have no page for), researches it with up to six web searches, writes the article, checks every claim against the sources it read, and publishes it to your site. Then it watches how the article does and uses that to pick the next one.

Will it make things up?

It is built not to. Every number in an article must appear in the source it links to, and every cited sentence is compared with the source's own text. If an article still fails after one rewrite, it is held for you instead of being published. You see the full fact check for every article.

Do I need Search Console?

No. New sites can skip it, and topics come from your site and from research on what buyers in your market search. Connecting it later makes the topic picks much sharper.

Where do the articles go?

We can host your blog on your own subdomain (one DNS record), add a blog to a site built with an AI builder through a prompt you paste, or publish to WordPress, Ghost or any endpoint through a webhook. Shopify, Wix and Webflow are coming.

Can I approve articles before they go live?

Yes. Choose Approve first and each article arrives by email with an Approve and publish button. Or leave it on autopilot.

Does it help with ChatGPT and Perplexity?

Articles answer the question in the first paragraph, cite sources inline and carry Article and FAQ structured data, which makes them easy for AI assistants to quote. Tracking when they do is coming soon.

How is Paperboy different from Outrank, Koala AI, Arvow, Surfer or Jasper?

Most AI SEO tools write and publish, and some read your Search Console. The difference is what happens before publishing: Paperboy saves the text of every source it reads and checks each number and citation in the article against it, on every plan, and holds an article for you if it still fails. Outrank sells human fact-checking as a $1,399 a month add-on; the others do not list an automatic check on their pricing pages (checked October 2026). If you are letting software publish under your name every week, that check is the feature to insist on.

Will AI-written articles hurt my Google rankings?

Google says appropriate use of AI is not against its guidelines; what it penalizes is content made mainly to manipulate rankings, however it is written. Paperboy is built on that line: real questions from your own search data, real sources, checked facts, and a sensible number of articles a week. We wrote up exactly what Google says in our blog.

What is GEO, and do I need it?

Generative engine optimization means being the page that ChatGPT, Perplexity, Claude and Google's AI Overviews quote when they answer. It rewards the same things good SEO does, plus answer-first writing and facts with sources. Every Paperboy article is written that way, and our free checker shows whether those tools can read your site at all.

How do I cancel?

On the Billing page, one button per site. Articles keep publishing until the end of the period you paid for. Everything already published stays yours.

See what it writes for your site