For agencies and web studios

Walk into the pitch with buyer evidence. Leave with copy your team can build.

Add a client’s website. TheLingo.ai reads what that client’s buyers say in public, sets the sellers apart, and returns the changes to make first: the line to replace, the text to paste, the posts behind it, and one brief your developer can build from.

You check the sources, the client confirms the facts, your team makes the changes. Nothing is published for you.

Neither button starts a report or asks for anything. A report runs only after you have a plan, have connected your own Apify and Claude accounts, and press start. Every step, in order, before you pay.

Where it fits in a client project

From the pitch to the build, on one client website.

1 · Pitch

Bring evidence, not opinion

Run the prospect’s website before the meeting. Show what their buyers keep asking, and which of it their site and their competitors leave unanswered.

2 · Kickoff

Set the starting picture

The first full report looks back a long way, one to three years in our examples, in the forums, communities and review sites chosen for that client’s market.

3 · Review

Check it before the client sees it

Read the fixes, open the posts behind the ones you plan to use, and collect the questions only the client can answer.

4 · Approval

The client confirms the facts

Answer those questions in the report and the copy is rewritten. A yes may be promised. A no is never claimed.

5 · Build

Hand over one file

Your developer or AI assistant builds from the brief. On the Agency plan the client files carry your agency’s name in place of ours, and the PDF and the page preview your logo.

One finding, start to finish

What you get, shown on one fix from the public sample.

The sample is a full report on callerwho.com, a phone system for home-service contractors. It read 7,616 public posts and replies; 837 were from real buyers, written by 676 accounts. Below is its first fix, copied from the report, in the order you would use it.

The finding

“Voicemail and a callback work fine”

Signal 97 of 100 · very strong75 posts and replies66 accounts10 places

A large group of established tradespeople say they are as busy as they want to be and that people worth working for leave a message or text. Some use the delay on purpose to filter out price shoppers.

Why it is first: buyers raise it, the site only touches it, and few sellers answer it.

The source

Where buyers said it

“People just leave a voicemail and I call them back at the end of the day.”
“I’m not exactly short of work so a £50 a month answering service would be pointless, when my phone contract includes voicemail.”

A quote is shown only when it is found word for word in its post, and it links to that post. All 75 are in the report. Not every one makes the point this directly, so open them and judge.

The change

Where it goes, and the text to paste

Where, on the home pageA new section directly after the section headed “You counted the missed calls. Then you called back at 8pm.” and before the section headed “What it says, and what it won’t.”
Add this textBusy enough already? Use it to pick your calls. Maybe voicemail and a callback at the end of the day work fine for you, and you’re as busy as you want to be. Then keep answering your own phone. Switch the AI off and it rings as normal. While it rings, the card shows the caller’s name, whether they own or rent, the home’s estimated value and the drive from your shop. Take the calls you want and let the rest go to voicemail, like you do today. Every voicemail is written out so you can read it between jobs.
Check with the client firstNothing for this fix. One of the sample’s four carries a question, the fourth: “Can an owner set a typical length for each job type and a travel buffer between jobs, and does the AI offer only times that fit both?” Until the client says yes, the brief tells the builder to skip that fix.

The site was read on November 19, 2026, each page to its end. A fix that replaces a line quotes it word for word from the page: the sample’s second replaces the paragraph that starts “It offers only times that fit your shop hours, each tech’s hours and the jobs already booked. Jobs outside your towns”. The new text is a draft, written from what the site already says.

The brief

What your developer or AI assistant gets

## Do these first

### 1. Voicemail and a callback work fine

Very strong signal, 97 out of 100: 75 real posts from 66 accounts in 10 places, 142 reactions. Buyers raise it and the site only touches it.

- **Page:** https://callerwho.com/
- **Where:** A new section directly after the section headed "You counted the missed calls. Then you called back at 8pm." [...]
- **Add this text:** Busy enough already? Use it to pick your calls. [...]
- **Also:** Add to the section headed "What owners worry about before they switch.": “Voicemail and a callback work fine for me.” [...]

### 4. Booking that allows for job length and drive time

Fair signal, 50 out of 100: 6 real posts from 5 accounts in 3 places, 10 reactions. [...]

**Only if the owner confirms:** Can an owner set a typical length for each job type and a travel buffer between jobs, and does the AI offer only times that fit both? Use the paste only if yes. If the owner has not said yes, skip this fix and list the question for them.

- **Page:** https://callerwho.com/
- **Find this line:** "It offers only times that fit your shop hours, each tech’s hours and the jobs already booked, so it won’t double-book."
- **Replace it with:** You set how long each kind of job takes and how much drive time to leave between jobs. The AI offers only times that fit both, [...]

One Markdown file: the page top to bottom, the lines to replace or the spot to add to, and the rule to promise nothing that is not in it. Download the sample’s brief.

Reading the full sample: which part matters first

The sample is long because nothing is held back. Read it in this order and the rest becomes the evidence behind it:

  • Fix these first. Four changes at the top, each with the page, the line or the spot, and the text. This is the deliverable.
  • One source. Open a quote’s link and read the post, so you know what a “real buyer” label looks like.
  • Check first. The questions for the client. The sample has 10; a fix that depends on one waits for its answer.
  • The brief. Download it from the top of the report and hand it to whoever builds.

What each report hands over

Findings your client can check, and files your team can use.

Every count is made by code from the posts, never by the AI, and every quote links to its post. The copy is a draft until the client has confirmed it.

  • The fixes to make first, with the page, the line to replace, the text to paste and the question for the client.
  • What buyers keep saying: problems, doubts and price talk, ranked, each with quotes and links, and whether the client’s site answers it.
  • The competition: the products buyers name, and each competitor’s own site read against what buyers raised.
  • Draft copy: home page openings and sections, four emails, four text messages, a press release, and replies for up to 40 forum threads worth answering.
  • A preview of the page with the new content, in the client’s own colours.
  • The downloads: the report as a PDF or Markdown file, the rebuild brief, and every post and lead as a spreadsheet.
  • Your name on the files. On the Agency plan, set your agency’s name and logo in Settings: the client files carry your name in place of ours, and the PDF and the preview your logo.

Before you pay

Every step between this page and a finished report.

There is no trial and no refund, so here is all of it first: what you pay us, what you pay two other companies, what you have to set up, and who does what.

  1. Pay for the plan You · $999 a year

    Choosing the plan opens “Create your account” (an email and a password), then Stripe’s payment page. The plan is charged once a year and renews until you cancel. All sales are final: no refunds, no trial.

  2. Open two provider accounts You · their sign-up

    Reports run on your own accounts with two companies: Apify, which collects the public posts, and Anthropic’s Claude Console, whose AI sorts the posts and writes the report. A Claude chat subscription does not include this; it needs an API key from the Console.

  3. Add both keys in Settings You · a few minutes

    Paste your Apify API token and your Claude API key. They are stored encrypted, used only to run your reports, and never shown again. Settings can test each key before you rely on it.

  4. Start the report You

    Enter the client’s website. Optional context makes it sharper: their best customer, competitors that must be included, the doubt they hear on sales calls, places their buyers talk, and words that must never be used. You choose how far back to look and how many posts to read in each place. Starting it uses one of your 15 website places.

  5. The research is billed to you Apify and Anthropic

    Each bills your own account for what the report used, with nothing added by us, and the report states what it cost. Three of our own runs: $10.17 looking back one week (775 posts and replies), $16.58 looking back a year at up to 50 posts a place (1,384), and $60.90 looking back three years at up to 1,000 a place (7,616). That last one is the public sample: reading it again (its findings on November 19, 2026 and again on November 20, and its competitors’ sites to the end of each page) added $38.91 of Claude. These are real runs, not a cap: a bigger search costs more.

  6. It runs Most: 25 minutes to 2 hours

    A very large search can take much longer. You can close the page. An email arrives when the report is ready, with a link to it and to each file. Two reports can run at the same time.

  7. You review it You

    Open the posts behind the fixes you plan to use. Whether a post is from a buyer or a seller is the AI’s reading of it, and the posts are there so you can check that reading.

  8. The client confirms Your client

    The report lists what only the business can answer. Until they do, the copy that depends on it stays in brackets or out of the page. Nothing a report writes is approved for publication by us.

  9. Your team implements Your developer

    Download the brief, the PDF and the page preview, and make the changes on the client’s site. TheLingo.ai publishes nothing, posts nothing and holds no login to the client’s site or to any forum.

Why a year, and not one report

A client’s buyers keep talking. The plan is priced for coming back.

There is no fee per report: run as many as you like on each client website during the year, and pay the providers for what each one reads. In client work that follows the calendar you already keep.

WhenWhat you runWhat it is for
Before a pitchA report on the prospect’s website.Evidence for the meeting. It takes one of your 15 places, so spend it on pitches you expect to win.
At kickoffThe first full report, looking back a long way: one to three years in our examples.The starting picture: problems, objections, competitors, and the first round of page changes.
Before a campaign or a launchA follow-up over the weeks since the last report.“What’s new” shows the buyer posts no earlier report had, competitors named for the first time, and how the fix list changed.
At a monthly or quarterly reviewA short follow-up. Our one-week run cost $10.17 in provider usage.Fresh objections for the next round of optimisation, with the posts to show the client why.
Between reportsThe weekly watch, switched on per website.It reads the same places every week and emails you the real buyers your reports have not heard. A quiet week sends nothing.

How often a re-run is worth it depends on how much that client’s buyers talk in public. The first report shows you: a market with little new talk gives a follow-up little to add, and a busy one changes month to month. Set the rhythm per client after the first report, not before it.

Pricing

The Agency plans, and what comes on top.

Two parts: the plan, billed by us once a year, and the research, billed to you by Apify and Anthropic on your own accounts for what each report uses.

Tick this box first. The plan buttons below need it.

Agency

$999 a year

$83.25/mo, billed annually

For client work, plus research usage.

  • 15 client websites, one per client
  • $66.60 per client website a year when all 15 are used, plus provider usage
  • Unlimited reports on each website for the year, the weekly watch included
  • Your agency’s name on the client files, and your logo on the PDF and the preview
  • Need more than 15? Agency Unlimited has no limit.

Agency Unlimited

$2,499 a year

$208.25/mo, billed annually

For agencies with more clients, plus research usage.

  • Unlimited client websites, one per client
  • Unlimited reports on each website for the year, the weekly watch included
  • Your agency’s name on the client files, and your logo on the PDF and the preview

The rules, so nothing surprises you later

  • On Agency, each client website takes one of the 15 places, and a prospect’s website you research for a pitch takes a place too. Agency Unlimited has no cap.
  • A website cannot be swapped for another during the plan year.
  • Every report needs your own Apify API token and Claude API key. Each provider bills you directly.
  • Cancel the renewal at any time in Settings. The plan keeps working to the end of the year you paid for, and payments are not refunded.
  • Reports download as PDF and Markdown files with every quote and link. They are yours to keep.

One business only? The Founder plan is $99 a year for one website or one product idea.

Choosing the plan opens “Create your account”, then the secure payment page. You connect your own Apify and Claude accounts in Settings before your first report. No report runs until you start one.

What it does not do

Four limits to know before you show a client.

It counts posts, not people.

Counts are of posts and replies. Where a report gives the accounts behind them it says so: the sample’s 837 buyer posts were written by 676 accounts. “Real buyer” is the AI’s reading of a post, shown with the post so you can disagree.

“Only you say it” means in the posts it compared.

A claim is listed as the client’s own when the AI finds it in none, or at most two, of the seller posts that make a claim. A report compares every one of them: the sample compared 1,766 of its 1,837 seller posts, the ones that make a claim. It is a finding about those posts, not about every competitor, so check a competitor’s site before you call anything unique.

It does not prove demand or promise a result.

Public conversations show problems and buying signals. They do not show who will buy from your client. Speak with prospects and test a real purchase decision.

It reads what is public and reachable.

Some places refuse or cannot be read. The report lists each place, how much was read there, and why a place could not be read, instead of showing zero.

Read the sample the way a client would. Then decide.