See which pages decide who AI recommends.
We ask ChatGPT, Gemini, Perplexity and Google’s AI your customers’ buying questions, eight times each. Then every page behind their answers, and whether you can get on it.
$650 for the first measurement. Report on day three. A walkthrough is included, if you want one.
Named Vendor A · share of all sampled answers · Sep 2026
165 of the 1398 answers named Vendor A. The bracket is the range, where the true figure very likely sits. The n is how many times we asked.
ChatGPT7.5% (5.0–11.2, n=280)
Gemini2.5% (1.2–5.1, n=280)
Perplexity20.7% (16.4–25.8, n=280)
Google AI Overviews19.4% (15.2–24.5, n=278)
Google AI Mode8.9% (6.1–12.8, n=280)
A real measurement of one mortgage software company. Its name and the names of the nine rivals measured with it are kept anonymous. The same 35 questions, 8 times each on every engine, 15–22 Sep 2026.
How to read the number
We show our math.
Ask an AI engine the same question twice and you get two different answers. So one sample is a guess wearing a suit. Every rate we report carries how many answers we sampled and the range the true number very likely sits in. If we cannot say that, we do not report the number.
The same measurement, shown two ways
This is how to read every number we print. We asked 1398 times. The engines named Vendor A in 165 of those answers. That is 11.8%. If we kept asking, the true figure would very likely land between 10.2% and 13.6%. The bracket is the range. The n is how many times we asked.
The pages behind the answer
An AI answer is built from a few pages
This is why the pages, not the answer, are the thing you can act on. Ask an AI engine what to buy and it answers with a few names, not a list of links. It builds that answer from a handful of pages, and those are the ones worth going after.
A staged example
The brands and the pages below are invented.
One answer looks like this:
- 1.TableBooks
- 2.LedgerPine
- 3.CountRoom
Is your brand named here?
Three names. Six pages. That is the whole competition.
How it works
Here is what happens between your questions and your report.
01 · The questions
You pick the questions. We ask each one 40 times.
Eight times on each of five AI engines. You get the question list back before we start, so you can cut anything that isn’t a real buying question.
Your question list, confirmed before the first run.
Mortgage software — buying questions
35 questions · 8 runs of each
ChatGPTGeminiPerplexityGoogle AI OverviewsGoogle AI Mode
- p01 What is the best CRM for mortgage loan officers?
- p02 Best mortgage CRM in 2026?
- p03 Which CRM do top-producing loan officers use to manage their pipeline?
- p04 I'm a mortgage broker. What CRM should I be using?
31 more questions in this set, 8 runs of each on every engine. From a real measurement, with the vendors’ names withheld.
02 · The run
We spread the asking over three days.
So no single day decides your number. Your report is ready on day three.
A count of answers back, on your brand page.
Run started 15 Sep 2026
Preliminary — ranges tighten as sampling completes.Paused until 16 Sep 2026, 21:34 UTC. This run’s samples are spread across 3 separated days, so one day’s conditions cannot define the measurement.525 attempted· 524 successful · 1 failed
- ChatGPT105/280
- Gemini105/280
- Perplexity105/280
- Google AI Overviews105/280 · 1 error
- Google AI Mode105/280
Day 1 of 3. 525 of this run’s 1400 answers are back, on ChatGPT, Gemini, Perplexity, Google AI Overviews, Google AI Mode. If an answer fails, we ask again. The first day of the same real measurement.
03 · The report
Every number with its range. Then the pages behind it.
Ordered most-cited first, each one graded by what it would take to get named on it.
One share link. No login for whoever you send it to.
Named Vendor A · share of all sampled answers · Sep 2026
Report #1 · next month can show a change of 4.2 percentage points or more
22 Sep 2026 · n=1398 · 5 AI engines · a real measurement, names withheld
The top of the report at full size, cropped to this column. The same real measurement.
Getting on the page
What it takes to get on a page
You can start with the pages that are open to you. Every page we could read carries one of three grades, and the grade says what it would take to be named on that page. Pages we could not read are marked Unknown and never counted as reachable (4 of these 25).
- Placeable12 of 25 pages · $0–500 · 7–90 days
- You can get on it yourself. Ask, submit, or pay a small fee. Community threads, independent comparison pages, review directories.
- Gettable3 of 25 pages · $250–3000 · 30–180 days
- Someone else decides. Pitch the editor, write for them, or pay for a spot. Major publications, single-product reviews, affiliate comparison pages.
- Locked6 of 25 pages
- There is nothing to buy into. A search engine’s own surface, a closed directory, a rival’s own page. Spend the effort where you can actually get on the page.
You get a money band and a time band for each grade. They run from the cheapest and quickest page of that grade among this example’s 25 most-cited pages to the dearest and slowest. Every one of those pages is in the table below.
You get that list, with the most-cited pages first. Getting onto one of those pages is how a brand ends up in the answers, and next month’s measurement is what shows whether it worked for you.
The proof
This is the report itself, not a picture of one, and it is what an agency hands a client. It lists every page the AI engines cited when they answered. Each row says who controls that page and what it would take to be named on it. It is a real measurement of one mortgage software company, taken 15–22 Sep 2026. Its name, the names of the rivals measured with it and their own pages are kept anonymous. Every number is as measured.
The finding
12 of the 25 most‑cited pages are placeable.
The 25 pages the AI engines cited most, out of 1,698 they cited in all. The count is how many times each page was cited across those 1,398 answers. The grade is measured, not guessed: ● placeable, ◐ gettable, ○ locked, ◌ unknown.
| # | Page | Reach | Cited | Who controls it | Est. cost | Est. time |
|---|---|---|---|---|---|---|
| 1 | mortgageadvisortools.com | Placeable | 639 | Comparison page — adjacent business | $50–500 | 14–90 days |
| 2 | ascendix.com/blog | Unknown | 545 | Unknown | — | — |
| 3 | maximizer.com/blog | Gettable | 526 | Single-product page | $250–1500 | 30–120 days |
| 4 | leadpops.com/guide | Placeable | 500 | Comparison page — adjacent business | $50–500 | 14–90 days |
| 5 | monday.com/blog | Gettable | 425 | Major publication | $500–3000 | 90–180 days |
| 6 | pipedrive.com/en | Locked | 378 | Competitor's own page | — | — |
| 7 | creatio.com/glossary | Unknown | 373 | Unknown | — | — |
| 8 | reddit.com | Placeable | 354 | Community | $0–150 | 7–45 days |
| 9 | Vendor J's own site | Locked | 330 | Competitor's own page | — | — |
| 10 | Vendor H's own site | Locked | 329 | Competitor's own page | — | — |
| 11 | reddit.com | Placeable | 302 | Community | $0–150 | 7–45 days |
| 12 | nutshell.com/blog | Placeable | 281 | Comparison page — adjacent business | $50–500 | 14–90 days |
| 13 | integrateiq.com/blogs | Placeable | 260 | Comparison page — adjacent business | $50–500 | 14–90 days |
| 14 | medium.com/@forsi | Gettable | 193 | Major publication | $500–3000 | 90–180 days |
| 15 | ascendix.com/blog | Unknown | 191 | Unknown | — | — |
| 16 | Vendor A's own site | Locked | 187 | Your own page | — | — |
| 17 | docspire.ai/blog | Placeable | 182 | Comparison page — adjacent business | $50–500 | 14–90 days |
| 18 | Vendor H's own site, page 2 | Locked | 180 | Competitor's own page | — | — |
| 19 | Vendor H's own site, page 3 | Locked | 164 | Competitor's own page | — | — |
| 20 | blog.fundmore.ai | Placeable | 164 | Comparison page — adjacent business | $50–500 | 14–90 days |
| 21 | softbliq.com | Placeable | 129 | Comparison page — adjacent business | $50–500 | 14–90 days |
| 22 | softwarefinder.com | Placeable | 128 | Comparison page — adjacent business | $50–500 | 14–90 days |
| 23 | reddit.com | Placeable | 127 | Community | $0–150 | 7–45 days |
| 24 | floify.com/blog | Placeable | 125 | Comparison page — adjacent business | $50–500 | 14–90 days |
| 25 | realestatebees.com | Unknown | 121 | Unknown | — | — |
Cost and time bands are estimates for a page of this type, not quotes. A vendor’s own site is locked to everyone else, so its address is withheld with its name.
Where the citations go
36.4% of the citations in these 1398 answers point at a placeable page. This counts citations, not pages, which is why it differs from the count above. Pages we could not read are 9.4% of them, and are never counted as placeable. The two shares split one total, so neither is a draw and neither carries a range. The share is context. The list above is the claim.
Who gets named
How often each engine cited any page
We never rank you, and we never score how you were talked about. 8 runs per question is nowhere near enough to say either one honestly.
Smallest period-over-period change this sample can detect: 4.2 percentage points. Anything smaller is indistinguishable from noise. Measured monthly, never weekly.
What we won’t do
Three things we won’t sell you.
We won’t rank you. Eight runs a question is nowhere near enough to order a list honestly.
We won’t score sentiment. Same reason. A number that shaky isn’t worth the confidence it would give you.
We won’t tell you a placement will move your number. Next month’s measurement is what shows that.
Made for agencies to run for their clients. A brand can run it for itself. It’s the same report either way.
For agencies
Sell it as yours.
You can put your own name on every report. Your logo at the top, your name in the footer, your color on the headline number. You send a link and your client opens it without an account or a password. The numbers, the ranges and the sample sizes stay as measured — the honesty of the report is not skinnable. You run the measurement. Your client reads your report.
Two invented agencies. Mapford has no agency clients yet.
Vendor A — Report #1
prepared by Harbor & Peak Digital · harborpeak.example
Vendor A — Report #1
prepared by Bright North Marketing · brightnorth.example
Plain answers
What do I actually get?
A report. It lists the pages the AI engines cite before they recommend businesses like yours. It says whether you are named on each one. Then it puts the pages worth going after in order, each with a cost band and a time band. Every number in it carries its range and how many times we asked. The sample report is a complete one, from a real measurement with the vendors’ names withheld.
How is this different from ranking on Google?
A Google ranking is a place in a long list of links. An AI engine shows no list. It answers with two or three names, built from a handful of pages it reads as it answers. Those pages are what we measure. It is a much smaller election, with different voters.
Do you guarantee ChatGPT will recommend me?
No. And be suspicious of anyone who does. AI answers change from one asking to the next. What we can show you is which pages the engines cite, whether you are on them, and which ones are worth going after.
Where do the numbers come from?
From asking, over and over. We put each buying question to each AI engine at least eight times. We record every page it cited. Then we report the rate with its range and how many times we asked. How we measure explains every step in plain English.
How often do you measure?
Once a month, never weekly. Ask an AI engine the same question twice in one day and the pages change about as much as they do a day apart. So a week of answers mostly measures the engine’s own randomness. A month is the window the research recommends. Every report also prints the smallest change it could have seen. In the sample report that is 4.2 percentage points. Anything smaller is reported as no change.
How fast do I get the first report?
Three days after the run starts. We spread the asking over three separate days, so no single day decides your number. Your report is ready on day three. A walkthrough is included, if you want one.
What does it cost?
A first measurement with a walkthrough of the report is $650. The agency license is $850 a month with 10 measurements included, so you can run one for each client. Pricing has the rest. There is no checkout and no free trial, because a measurement is only worth buying if it fits your business.
How do agencies use it?
As their own product. Every report is white-label: your logo, your name, your color. Share links open without an account. Agencies run the measurement for each client and deliver the report under their own brand.
One measurement
Start with one measurement.
35 buying questions, 8 times each, on 5 AI engines. That’s 1,400 answers, every number with its range, and the pages you can reach.
$650 for the first measurement. Report on day three. A walkthrough is included, if you want one.