Skip to content
sourcesignal.ai
Outreach recordpages 443spent $3,449AI answers read 24,699Jul 20 to Nov 11, 2026

We spent about $3,400 to get named by AI. Here’s the receipt.

AI engines answer buying questions from a handful of pages. So we ran an experiment on ourselves. sourcesignal asked 443 of the pages they cite to add a business, paid when the price fit, and watched what the engines said next.

For our older brand, the answers reading those pagesAn AI answer that lists the page among its sources: the answers that read the page. went from naming it 4% of the time to 51%. For our newer brand, 571 answers later: not once. Here is how it went, door by door, and what we think split the two.

01 / Knocking

We knocked on 443 doors. One in five opened.

Every door was a page an AI engine cites for a buying question, most of them best-of lists. Where we wrote mattered most: the address a site printed itself answered nearly three times as often as one a lookup toolAn email-finding service: it searches a whole domain, or a person’s name at a domain, for an address. dug up, 29% against 11%.

Answered
9020% of 443 pages
Named a price
5561% of those that answered
Said yes
15verified live on the page
Never answered
35380% of the pages

The 443 pages, one field of pixels sorted four ways: each reading lights its share of the pages. Pick a reading to hold it still.

See the numberswhere we wrote, how fast they answered, and which pages did

Where we wrote

Pitches to an address printed on the site were answered 29% of the time; pitches to an address a lookup toolAn email-finding service: it searches a whole domain, or a person’s name at a domain, for an address. found by searching the whole domain, 11%. The gap holds inside each business. Outreach advice often says to find a named person, but among printed addresses a shared inbox such as editor@ answered as often as a person, 29% each.

  • Printed on the site: 29%, 45 of 155 pitches
  • A named person, found by a lookup tool: 20%, 16 of 82 pitches
  • Guessed from the site’s address pattern: 18%, 9 of 51 pitches
  • A lookup tool’s search of the whole domain: 11%, 21 of 194 pitches
Pitches answered, by where the address they went to came from. A pitch counts once, however many follow-ups it had.
  • Business A

    Printed on the site: 34%, 18 of 53
  • A lookup tool’s domain search: 14%, 11 of 81
  • Business B

    Printed on the site: 36%, 17 of 47
  • A lookup tool’s domain search: 15%, 7 of 48
  • Business C

    Printed on the site: 21%, 10 of 47
  • A lookup tool’s domain search: 5%, 2 of 43
The same comparison inside each of the three businesses with the most pitches.
  • A shared inbox, like editor@: 29%, 26 of 90
  • A named person: 29%, 19 of 65
Printed addresses only: an address whose first part is a role word (info, hello, editor, press) against one that is not.

How fast they answered

Answers came fast or not at all: a quarter within 41 minutes, half within 16 hours, three quarters within about four days. Follow-ups still count: 26 of the 91 replies came to a second or later email.

Of the 91 pitches that got an answer, a quarter got it within 41 minutes, half within 16 hours and three quarters within about four days.

Time from the first email to the first answer, for the 91 pitches that got one, on a clock from ten minutes to two weeks.

Which pages answered

Between Jul 20 and Nov 11, 2026, sourcesignal sent 483 pitches and 1,051 emails to 370 sites, for five businesses: our two brands, a customer on a trial and two other projects. Best-of lists, 362 of the pages, answered 21% of the time, comparison pages 32% and editorial articles 12%; none of the seven directories answered.

  • Best-of lists: 21%, 76 of 362 pages
  • Comparison pages: 32%, 7 of 22 pages
  • Editorial articles: 12%, 5 of 42 pages
  • Review pages: 14%, 1 of 7 pages
  • Directories: 0%, 0 of 7 pages
Pages that answered, by the kind of page the platform read it as. Kinds with fewer than seven pages are left out.
n = 443 pages, 483 pitches, 1,051 emails370 sites, 5 businessesn = 482 pitches with a known address sourceanswered: a real reply, not a bounce or an out-of-office noticeprinted: shown on the publisher’s own site18 pitches bounced

02 / The price

Most doors that opened had a price. Answer it, and it can come down.

Six in ten pages that answered named a fee, a median of $250 to start. A third of the quotes we answered fell, to about half; none we left alone moved. Almost every yes came this way: 13 of our 15 placements followed a price.

First quotes from 49 publishers: a median of $250, the middle half from $130 to $610, from $49 to $1,795.

median first quote $250

Each publisher’s first quote, one dot each, 49 publishers, on a log dollar scale. Hover a dot for its quote and what it was for.
See the numberswhat came down, what paying more bought, and what set the price

What came down

Of the 37 quotes we answered, 12 fell, to a median 54% of the first figure, and 3 rose; none of the 19 we left unanswered moved. A page that answered with a price added the business 24% of the time; a page that answered without one, once in 35.

12 of the 37 quotes we answered came down, to a median 54% of the first figure, between 26% and 75%. $1,700 to $450; $690 to $300; $650 to $455; $637 to $319; $610 to $460; $585 to $195; $575 to $288; $500 to $350; $350 to $200; $250 to $100; $200 to $120; $200 to $150.

The 12 quotes that came down, first figure to last, highest first. All 12 came after we wrote back.
  • Answered with a price: 24%, 13 of 55 pages
  • Answered without one: 3%, 1 of 35 pages
Pages that answered, by whether the answer named a price, and the share that then added the business.

What paying more bought

Cheaper quotes said yes more often, 35% under $150 against 13% from $400, but mostly that is our ceiling. Our older brand paid up to £400, and none of the 12 quotes above it ended in a placement; our newer brand paid up to $600 and countered 3 of the 5 quotes above it into one, $1,700 down to $450. Placements went live a median 8 days after the pitch.

  • Under $150: 35%, 6 of 17 pages
  • $150 to $399: 27%, 4 of 15 pages
  • $400 and up: 13%, 3 of 23 pages
Pages that named a price, by their first quote, and the share that then added the business. Our ceilings decide much of it: £400 a page for the older brand, $600 for the newer.

What set the price

Each niche had its own going rate: a median first quote of $500, $250 and $175 for the three businesses with the most quotes. Domain authority barely set it: across the 37 priced pages with a rank, price and authority correlate at 0.25, and at 0.09 inside our older brand’s niche. Middling sites asked the most, a median of $513, against $275 for high-authority ones and $140 for low.

37 pages that named a price and have a domain rank. The rank correlation of price and authority is 0.25. Low-authority sites asked a median $140, middling ones $513 and high-authority ones $275.

The 37 priced pages with a DataForSEO domain rank: authority across, the first quote up, on a log dollar scale. Accent: the pages that then added the business. The band is our two price ceilings.

When it came, and what for

In 44 of 56 priced threads the price came in the first reply, a median 20 hours after the first email. The middle half of first quotes ran from $130 to $610, the range from $49 to $1,795. A sponsored post was quoted highest, a median of $400, against $250 for a placement fee.

  • A sponsored post: $400, median of 12 quotes
  • A placement fee: $250, median of 31 quotes
  • A link in an existing page: $225, median of 8 quotes
Median first quote by what the publisher was selling, in the three kinds with the most quotes.
n = 49 publishers’ first quotes, 56 priced threadsdollars at October’s rates: £1 = $1.30, €1 = $1.15quotes, not paymentsanswered: we wrote back after the pricesaid yes: the business verified live, from a tracked pitch

03 / The wait

Then we waited. Three weeks later, the answers had changed.

We followed the answers on 16 pages that added our older brand, eight weeks either side of the day each went liveThe day the platform first found the brand on the page. Some pages may have added it a little earlier.. For a week, little changed. Then the answers reading those pagesAn AI answer that lists the page among its sources: the answers that read the page. named the brand more and more: 10% of the time before, 47% by week three. Nobody was pushed out to make room. An AI answer is a list with room in it.

Answers citing the page named the brand 10% of the time in the eight weeks before its placements; 13% in the first six days after; 19% a week after; 29% two weeks after; 47% three weeks after; then 30% by the eighth week. Other answers to the same questions stayed between 9% and 15%.

One brand, 16 placements with five or more citing answers on each side, answers grouped by week from the day each placement was verified live (week 0 is the first six days after). Accent: answers citing the page; gray: other answers to the same questions.
See the numbersweek by week, engine by engine, and the room in the answer

Week by week

For our longer-running brand, 16 placements had enough answers to read: 19% a week after, 29% two weeks after and 47% three weeks after, settling near a third. Answers to the same questions that do not cite the page stayed between 9% and 15%. One page’s answers went from naming the brand in 0 of 34 to 39 of 48, starting 5 days after it went live. Our last paper could not measure a lift from one page across client projects; watched before and after, the answers that read the page changed.

Engine by engine

ChatGPT went from 14% to 45% and Google AI Overviews from 20% to 45%. Claude went from 2% to 10% and Gemini from 5% to 10%, both on fewer answers. The platform’s own flag, which needs the brand’s full name, gives 9.8% to 31.6% across all four, against 10.5% to 12.7% in the rest.

Share of answers naming the brand before and after its placements, in answers citing the page and in the rest: All four engines, citing the page 14% to 37%, other answers 12% to 14%; ChatGPT, citing the page 14% to 45%, other answers 14% to 20%; Google AI Overviews, citing the page 20% to 45%, other answers 20% to 25%; Claude, citing the page 2% to 10%, other answers 3% to 6%; Gemini, citing the page 5% to 10%, other answers 6% to 6%.

One brand, 16 placements, 56 days either side. The brand counts as named when its name, or the first word of its website, is in the answer’s text.

Your own site comes along

Answers citing the page cited the brand’s own website 19% of the time before and 36% after. A placement can bring your own site into the sources.

  • Answers citing the page

    Before: 19%, 235 of 1,215
  • After: 36%, 508 of 1,393
  • Other answers to the same questions

    Before: 14%, 1,625 of 11,995
  • After: 16%, 4,088 of 25,680
Answers that list the brand’s own website among their sources, before and after its placements.

Room for one more

If an answer names a fixed number of brands, getting in should push someone out. It did not: answers citing the page named 0.7 more competitors afterward, the same rise as every other answer to those questions (0.7), and added the brand on top. They name about seven businesses, and the brand was named alone once in 440.

Before the placement

AI answer

6.6 competitors named per answer; the brand in 10%

After

AI answer

7.3 competitors named per answer; the brand in 32%

Drawn: an answer citing the page as a ranked list, competitors in gray. The counts under each are measured; the names and their order are a drawing, not one particular answer.
n = 16 placements, one brand2,608 answers citing the pages, 37,675 othersChatGPT, Google AI Overviews, Gemini and Claudeverified liveThe day the platform first found the brand on the page. Some pages may have added it a little earlier.: the day the platform found the brand on the pagenamed: the platform’s flag on the brand’s full nameown site: the brand’s website among the answer’s sourcescompetitors: the brands tracked for the brand’s questions; the list grew over the period, for both groups alike

04 / The receipt

One brand went from 4% to 51%. The other got nothing for $2,786.

Same playbook for both: pitch, answer the price, pay under a ceiling. Of the placements this budget bought, the older brand’s lift came from the 3 old enough to read: two free and one a $100 listing. The newer brand’s pages have been read by 571 answers since, and not one named it. The clearest difference: the engines already used the older brand’s name. A listing amplified it. It has not yet introduced a new one.

Our older brand: 5 paid placements for $663 and 2 free. On the 3 old enough to read, answers citing the page named it 4% of the time before and 51% after. Our newer brand: 8 paid placements for $2,786; 571 answers have cited those pages since, and none named it. In all, $3,449 spent against $4,570 first asked.

Spent: the last price named in each thread that ended in a placement; the record holds quotes, not invoices. What came back: answers citing the page that name the brand, 56 days either side of each placement going live, in the platform’s daily checks.
See the numbersthe spend, the brand the engines did not know, and what one page can move

What we spent

We gave ourselves a rule: counter anything above £400 a page for the older brand and $600 for the newer, and pay when a price fits. In all we paid for 13 placements, $3,449 by the last price named in each thread, against $4,570 first asked: answering the quotes took $1,121 off. The older brand’s $663 bought five placements, two more came free, and the other 4 are still too new to read.

The brand the engines did not know

The newer brand got onto nine cited pages from Oct 13. The 571 answers citing them since named 7.1 competitors on average and never the brand, while the older brand’s rise showed within two weeks.

571 answers cited pages that list the brand after it was added. None named it; they named 7.1 competitors on average.

Every answer that cited one of the newer brand’s pages after it was added, one square each, 7 placements with enough answers.

What one page can move

After a placement the page sat in about one answer in 20 for its question, so a 22-point rise there is worth about one point on the question. Across 29 questions the brand was named in 11.1% of all answers in the four weeks before and 10.6% in the four weeks after. Moving a question takes most of the pages it cites, which is where Links get you cited. Mentions get you named. left off.

After a placement the page sat in about one answer in 20 for its question. A rise of 22 points in those answers is worth about 1 point on the question.

Drawn: twenty answers to one question, one citing the page. The share and the rise are the older brand’s 16 placements, after each went live.
n = 13 paid placements, 2 freeceilings: £400 and $600 a pagethe older brand’s other 4 placements: 7 to 15 days oldthe newer brand: 9 placements from Oct 13questions: 29 with a placement four weeks old, 28 days either side

05 / What it means

Getting onto the pages AI reads works, if the engines already know your name.

01Write to the address the site prints
Not the one a tool finds. A shared inbox answers as often as a person.
02Expect a price, and answer it
Most doors that open have one, and only an answered price comes down.
03Give it three weeks
The answers that read the page start to move after a week.
04Get known before you get listed
A listing amplifies a name the engines already use. Being written about may have to come first.
05Think in questions, not pages
One page is about one answer in 20. Moving a whole question takes most of the pages it cites.

sourcesignal finds the pages AI cites for your buyers’ questions and the address each page prints, drafts the pitch and sends it from your own mailbox once you approve it, then watches the answers before and after you are added.

Questions worth asking

What the record suggests and does not settle.

Each has a number behind it, and none is proven. Open one to see it.

  • 01Is a line in an AI answer for sale, at about $150?
    13 of the 15 placements came after a quoted price, and the last price named on those pages had a median of $150. On our older brand’s 7 placements from tracked pitches, 5 of them on priced pages, answers citing the page went from naming it 7% of the time to 51%. We hold quotes, not payments.
  • 02Does paying more buy more?
    The biggest lifts in the record came from two free placements and one at $100; the newer brand’s $2,786 has bought none yet. Too few placements to say, which is why it is a question.
  • 03Are you paying for authority?
    Barely, by our record: price and domain authority correlate at 0.25 across 37 pages, and at 0.09 inside our older brand’s niche. Search keywords track the price more closely inside each niche, about 0.5.
  • 04Do publishers know what their pages are worth to AI?
    Across the record, a page’s price had nothing to do with how often AI cites it (a rank correlation of 0.05). Inside one niche the most-cited pages cost three times as much, $610 against $200: the price seems to follow the niche first and the AI attention second.
  • 05Are the pages AI loves the easiest doors?
    The most-cited third of the pages answered more often (24% against 18%) and gave 10 of the 15 placements. It does not hold inside every business, so we ask it rather than claim it.
  • 06Which engines read the page, and which remember?
    ChatGPT and Google AI Overviews moved by more than 20 points; Claude and Gemini by 5 to 9, on fewer answers.
  • 07Are smaller sites the better doors?
    For our two brands, smaller sites answered more often: 28% against 21%, and 38% against 17%. The customer’s pitches saw the opposite.
  • 08Is the third email worth sending?
    It earned replies about as often as the second: about 4% of the pitches still waiting, against 5%.

Method

Where the record comes from

Every pitch sourcesignal sent between Jul 20 and Nov 11, 2026, read from the production database, and 24,699 AI answers from four engines. A before-and-after on one brand, not a controlled experiment. No page, business, person or address is named on this page.

Measured November 12, 2026 from the production database, read only. Figures describe Jul 20 to Nov 11, 2026; the record keeps growing.

Read the full method
The record
483 pitches to 443 pages on 370 sites, for five businesses. Two are our own brands, one is a customer on a trial, with their agreement, and two are other projects in our workspace. The platform’s reply reader marks bounces, out-of-office notices and any price a reply names; it read 267 of the 268 replies.
Answered, priced, placed
Answered: a reply after the first email that is not a bounce or an out-of-office notice. Priced: a reply the reader marked with an amount, each publisher’s first quote, in dollars at October’s rates (£1 = $1.30, €1 = $1.15). Said yes: the business verified liveThe day the platform first found the brand on the page. Some pages may have added it a little earlier. on the page and matched to a tracked pitch, 15 pages. Another 26 placements, logged by hand outside these pitches, stay out of the funnel.
The answers
For the two brands with placements, every answer the platform collected from ChatGPT, Google AI Overviews, Gemini and Claude for their tracked questions: 24,699 answers. A brand is named when the platform’s flag finds its full name; as a check we read the text for the name’s words in order, or the first word of the brand’s own website, and the engine figure uses that wider reading.
Before and after
For each placement, the answers to the questions its page is cited for, 56 days either side of the day it was verified live, split by whether the answer cites the page. A placement counts once it has five or more citing answers on each side: 16 for the older brand, 7 for the newer. The question-level reading takes 28 days either side, for placements at least four weeks old and questions with ten or more answers on each side.
The budget
Our two brands ran one fee policy: counter anything above a ceiling, £400 a page for the older brand and $600 for the newer, and pay when a price fits. Spent is the last price named in each thread that ended in a placement: $3,449 in all, or $2,890 at the lowest price named in those threads. The record holds quotes, not invoices. Authority is the latest DataForSEO domain rank (0 to 1,000) of the 37 priced pages that have one.
What this shows
Some pages may have added the brand before we verified them, which blurs the weeks just before; answers citing the page were already rising then, and other answers rose too. The newer brand’s placements are recent. Rates pooled across the five businesses can reverse inside one, so what holds only pooled is asked above, not claimed.

Your questions, your doors

See which pages answer your buyers’ questions, and who to ask.

Every finding above starts from the pages AI engines cite. Ask three of your own questions and the free report runs each on four engines: how often you are named, which competitors are named instead, and the pages each answer cites. Those are the doors worth knocking on.

Three questions your buyers ask, in their words

A question a buyer asks before choosing, not your brand name. One is enough to start; the free report takes up to three.

How the numbers are made