Answermend / about
About Answermend
We publish what
we measured.
And what we didn’t.
Answermend exists to answer one question: when a customer asks an AI assistant the question you want to be the answer to, does it name you? And when it doesn’t — what has to change on your site so that it might.
Measurement is the product. So the only thing worth putting on an about page is how we measure, what we refuse to claim, and where the boundary around it sits. All three are below, in that order.
7 of 8
Named in
at least one measured answer
3
Cited
page fetched + linked
4
Mentioned only
from model memory
1
Not measured
excluded from counts
| Prompt | ChatGPT | Gemini | Named instead |
|---|---|---|---|
| Best review software for contractors | cited | cited | — |
| Top review management providers | mentioned | missing | competitor-one.com |
| How much does review software cost? | missing | mentioned | competitor-two.com |
| Northlight vs alternatives | not measured | cited | — |
Who this is for
Small enough to notice AI
skipped you. Near enough to
the site to fix it.
The person this is built for runs marketing at a growing brand, or runs the brand, or runs an agency looking after a handful of them. They already have a site that ranks somewhere. What’s new is that a share of the questions that used to arrive as clicks now get answered before the click, and nobody in the building can say whether the brand appears in those answers or not.
Step one — see
Run the checks, and find out whether the answers name you
Four things can come back for any one question, and they are not degrees of the same thing.
cited mentioned missing not measured
Step two — mend
Get the fix as a file, not as a recommendation
The output of a gap is something you paste in, or forward to whoever built your site.
FAQPage schema block llms.txt entry answer section on the closest page
On the find side, three site checks — Crawler Checker, Readiness Score and Render Check — each fetch your site from our server and hand back a real result about your own pages. On the mend side, the four generators turn what you type into a file you can paste in, and the deeper fix layer is what the paid plans are for. The scope panel below draws the boundary around both halves.
It is deliberately not several other things. Not an enterprise platform — no SSO, no procurement track, no security questionnaire; that is an exclusion, not a roadmap item, and a buyer who needs those should buy elsewhere. Not a rank tracker or a keyword database. And not competing on how many engines get tracked: that is a well-funded arms race, and breadth is already being given away free inside tools most people already pay for.
The distinction underneath everything
Mentioned is not cited.
That gap is the product.
A brand can be named in an AI answer without the model ever fetching the site. That sounds like a technicality. It is the difference between two entirely different problems, with two entirely different fixes.
mentioned
Named from memory
The model produced your brand name out of what it absorbed during training and from whatever else it read along the way. Nothing of yours was fetched. Your site could have been offline for a year and the sentence would read the same.
Good news, up to a point: something out there taught the model you exist. But you have no control over the version of you it remembers, and no link back.
Fix lives off-site: what other sources say about you, and whether the model is certain which entity you are.
cited
Your page was fetched as a source
The engine went and got a page, used it, and pointed at it. That is the only state that proves your content was read rather than remembered — and the only one that can send a person to you.
It is also the state that fails silently. A blocked crawler, a page that only assembles itself in JavaScript, an answer buried three scrolls down a page — any of those and you are unquotable while looking perfectly healthy in Google.
Fix lives on-site: crawlability, rendering, and whether the answer is liftable in one piece.
Report one of these as the other and every number downstream is wrong — a brand that looks well-covered is actually invisible to the crawler, or a technical problem gets treated as a PR problem. So the two never merge, anywhere: not in the report, not in a score, not in a summary email.
The four states
Four states, and one of
them is not a zero.
Every check we report comes back as exactly one of these. The wording is fixed, because three of them get quietly rounded into each other by tools that would rather show a full grid.
cited
Your site was fetched as a source.
The engine read a page of yours and pointed at it.
mentioned
Named from model memory only.
Your brand appears. Your site was never fetched.
missing
Not in the answer.
The check ran, and you weren’t there. A real, measured absence.
not measured
The check didn’t run — never shown as zero.
A missing key, a failed call, a rate limit, an engine with no programmatic route.
| Engine | State | Runs that returned | Reason, when there isn’t one |
|---|---|---|---|
| ChatGPT | cited | 1 of 1 | — |
| Gemini | not measured | not applicable | rate limited — left out of every count, never counted as a nought |
The last one is the rule that costs us something. A tool that shows not measured as a zero looks more complete, scores lower for you and higher for whoever it did manage to check, and nobody ever finds out. It also makes a broken integration on our side look like a competitor beating you — which is the single most expensive kind of wrong a measurement product can be. So an unrun check says it didn’t run, and where an engine has no API we say it isn’t measurable rather than scraping something and calling it data.
The limits
What we won’t claim.
Each of these is a line we could cross to make a page convert better. They are written down so that crossing one has to be a decision rather than an accident.
-
Never
That we can guarantee a mention, a citation, or a score
Nobody controls what a model says about a brand — not us, not anyone selling you the opposite. What can be guaranteed is the deliverable and the measurement: the schema file, the crawler fix, the drafted page, and an honest reading of what happened after. Anyone promising the outcome is selling weather control.
-
Never
A single check dressed up as a measurement
Ask the same engine the same question twice and you can get two different answers. That means one check is a screenshot, not a number — useful as an illustration, worthless as a baseline, and dangerous the moment somebody puts it in a board deck. Anything we present as a trend has to have been measured more than once.
-
Never
“Runs in your browser, never leaves your machine” — on a tool that fetches your site
That line is true of the generators here: they only transform text you typed, so nothing is sent anywhere and you can confirm it in your own network tab. It is false of every checker that takes a domain. A browser cannot read another site’s
robots.txt,llms.txtor rendered HTML — cross-origin responses are unreadable without headers those files never carry — so that work happens on a server. Wearing the privacy line anyway would be the cheapest lie available, and someone always reads the policy. -
Never
A client name, a logo wall, a testimonial, or a “trusted by” number
None exist yet. When one does, it appears here with the customer’s permission and not before — and a count only appears once it is a number counted from the database rather than an adjective.
-
Never
Your scan data in anything public
A scan record holds your brand, your competitors, and precisely where you are losing. It does not go into a case study, a blog post, or an aggregate statistic without your explicit, separate opt-in — separate meaning its own decision, not a clause folded into terms you accepted to sign up. It is not used to train models.
Scope, plainly
What this does,
and what it will not.
The same panel sits on the about page and the contact page, because the same question arrives through both. Read the right-hand column first — it is the more useful half. Everything in it is a decision rather than a gap, so if you need one of those things, buy it somewhere else instead of waiting on us.
| Area | What Answermend does | What it deliberately does not do |
|---|---|---|
| What gets measured | Whether an AI answer cites your page, names you from memory without fetching anything, or leaves you out. Four states, kept apart — and not measured is never rounded down into a zero. | No single blended visibility figure, and no position number. An AI answer has no positions in it, so a rank would be a number we made up. |
| Which assistants | The free scan asks ChatGPT and Gemini. Perplexity is on the paid plans. | Not a breadth race. Where an engine has no public API — Google’s AI Overviews — it is named as unmeasurable rather than scraped and dressed up as a measurement. |
| What you get back | The fix as a file: a schema block, an llms.txt entry, an answer section for the page closest to the question. Something you paste in, or forward to whoever built the site. | No audit deck, no retainer, and nothing that edits your website for you. Nothing there changes unless you choose to paste something in. |
| Who it is for | One brand at a time — the person who owns it, the marketer who answers for it, or the agency looking after a handful of them. | Not an enterprise platform. No SSO, no procurement track, no security questionnaire. That is an exclusion, not a roadmap item. |
| What it costs to look | The generators and the site checks take no card, no account and no email. The result appears on screen. | There is no checkout on this site and no card is stored anywhere. Nothing renews, because nothing starts. |
| Your scan data | A scan record holds your brand, your competitors and precisely where you are losing. It stays yours. | It never goes into a public page, a case study or an aggregate without your separate, explicit opt-in. No model is trained on it. |
Why put that on an about page.
Because at this stage the only real asset here is that the numbers can be argued with. There is no dataset advantage, no head start, no logo wall. What there is: a product built to tell you when it didn’t manage to check something, to keep two facts separate that are easier to merge, and to hand over the fix as a file rather than as advice.
The day a not-measured gets rounded up to a zero to make a report look fuller, there is nothing left here worth buying. That is the whole reason the rules above are written down in public rather than kept in a document nobody outside can read.
An about page is also where a company usually borrows credibility it has not earned — a founding year that implies a decade, a “team” that is one person and a contractor, a customer count with a plus sign after it. Given that the argument on the rest of this page is that our numbers can be trusted, none of that appears here.
Built and run by one person: Saad D., Founder — Houston, TX, United States. Answermend is a product of AdModifier LLC, 30 N Gould St #40766, Sheridan, WY 82801. Write to hello@answermend.com.
Start with something that costs nothing.
The four generators are open — no signup, no email. So are the three site checks: Crawler Checker, Readiness Score and Render Check each fetch your site and return a real result.