You cannot see that conversation, and your own data shows only part of it. Headwater tests how the assistants answer your buyers' questions, finds where demand in your category is moving, and writes source-backed content under your own claims rules.
Plain answers first
AI visibility is whether an AI assistant such as ChatGPT, Perplexity or Gemini names your brand, cites your site, or recommends you when someone asks a buying question. It is measured by asking the assistants a fixed set of questions and counting the answers. The result shows how often you appear in the captured answers, not in every conversation buyers have.
Answer engine optimization (AEO) means publishing clear, well-sourced content for AI answers and measuring whether those answers change. You will also see it called generative engine optimization (GEO), AI search optimization, AI SEO or LLM SEO. The names are new and people use them loosely; they all describe the same work.
Headwater is a service Suggestic runs for you. We test a fixed set of your buyers' questions on the assistants before and after publishing, read where demand in your category is moving, and publish source-backed content on your own site under your claims rules. We do not assume the assistants will use the pages; we measure whether they do. Every count comes with the total it was counted against and the date.
What you can buy
Headwater is an operated service, not a platform your team logs into. Start with the part that answers the question in front of you.
Repeated, on the same questions. We put a defined set of your buyers' questions to ChatGPT, Perplexity and Gemini, plus Google's AI Overviews on live searches, and score whether you are named, how often, how you are described, and which competitor is picked instead. You choose the assistants for the study. We can add any assistant we can reach through a software connection or a live web page.
What you get
Weekly counts, from the sources you choose. We count demand signals across search, video, social and community, check them against each other, and rank them. In community threads we read for presence and tone, and every quote is checked against the thread. Where a category has research and product launches behind it, we read those too. We can use any source that gives access to its data, including websites, licensed sales or social-listening data, and your own internal systems.
What you get
Sources first, then the draft. We find and grade the sources before a word is drafted, write in your voice, check the draft by code against your approved claims and banned terms before a person reviews it, and publish it on your own site in a form crawlers can read. We can also prepare content for Amazon listings and the other places you sell. The output can take many shapes: articles, listicles, answer pages, product detail pages (PDPs) and other formats you need, agreed when we set the scope.
What you get
How the three are made
Code does the work and keeps the record. A language model is used only where judgment or wording is needed, and what it produces is checked by code against the rules you set before anything ships. Every review decision is written down, including what the review turned down.
How the work is kept honest
Every vendor in this category can show a chart. The question is what the chart counts, over what, and since when. These are the rules every Headwater figure follows.
Repeating a calculation on the same saved inputs gives the same result, and the inputs are kept with the work. New assistant answers and generated text can differ each time, so we save them separately and say so.
Every figure traces to its source. The model behind each captured answer is recorded on the answer, so a change in the model reads as a fact, not a guess.
Your data runs alone, in its own environment. You own the output. Nothing trains a shared model.
Limited evidence is flagged as limited. A trend you named before the run is reported with its rank whether or not it came back. Comparisons are reported whichever way they fall.
We do not control what an assistant says. We commit to the work: measure, find, write under your rules, publish, measure again, and report what moved.
A comparison page we never touch helps you judge whether the work contributed to a change. On its own it does not prove cause, and we do not say it does.
Finding a trend you already know is a calibration check. It is not evidence that we saw it before your sales did, so we do not quote lead time.
If you already run an AI-visibility platform
Platforms such as Semrush's AI Visibility Toolkit, Profound and Peec AI do a real job. They track how a brand shows up across the answer engines, and some also draft content with an approval step. Headwater is not sold against that. It is a different kind of thing: a team that investigates, decides, writes and measures, and leaves you owning the result. The honest way to compare is to put the same five questions to every provider you are considering. Here are Headwater's answers.
Put the same five questions to any tool or service you are comparing and read the answers side by side. You can keep a platform you already have and buy only the parts you need. If you do not have one, Measure can give you that reading. Platform details are from each vendor's own site, November 2026.
How we start
Choose a demand briefing, an assistant-answer measurement, or both. A briefing is a ranked list with the evidence behind every line, plus the check on a trend you already know. A measurement comes with the question set, the captured answers and the run conditions. Both end in one recommended next action. Write is set up in the build stage. If you choose Write alone, we agree what to write at scoping.
Only the parts you chose: demand reading tuned to your categories, your claims rules written into the review, publishing on your site, and tests built from your buyers' questions. The pages we will update are named separately from the pages we leave alone.
We operate and extend it. You choose how often briefings and measurements run, and each one includes its settings and its review record.
After an assessment you own the briefing and its exports; the public and licensed sources behind it stay under their own terms. After a build you own what was built for you: the vocabulary and rules as written files, the content, the captured answers and the measurement history, all exportable in open formats. The engine that runs it is ours, licensed to you for the term.
The rules you already live by: the claims you can make, the words you cannot use, and who signs off. Access to the data you already license, if you want it read alongside the public sources.
Common questions
We write a fixed set of questions your buyers would ask, lock it, and put it to ChatGPT, Perplexity and Gemini, plus Google's AI Overviews on live searches on a schedule. Each answer is stored with the question, the assistant, the model that served it and the sources it cited. We then count the answers that name you, cite your site, or recommend you, and report each count with the total it was counted against and the date. The result shows how often your brand appears in the captured answers to those questions, not how often it appears in every conversation buyers have.
SEO focuses on visibility in search results. AEO and GEO focus on how a brand appears in AI-generated answers. The two newer names describe the same work: clear, well-sourced pages on the questions buyers ask, published on your own site in a form crawlers can read, and a measurement to show whether the answers changed.
ChatGPT, Perplexity and Gemini today, plus Google's AI Overviews captured on live searches. The set is not fixed: any assistant we can reach by API or on a live page can be added to a study, and the question set is written around the assistants your buyers actually use. The model behind each answer is recorded, so when a provider changes models the change shows in the record rather than hiding inside the trend.
On the cadence you set. A single run is a snapshot. Repeating the same questions every few weeks turns it into a line, and that line is what lets you read a change against the comparison page.
No, and we would be wary of anyone who does. The assistants decide what they say. We commit to the work and to reporting what moved, whichever way it falls.
No. Keep it. Sense finds what to write about, Write produces it under your rules, and Measure tests how the assistants answer your buyers' questions. The three parts are bought separately or together, so Measure can be left out if your platform covers it, or added if you do not have one. The scoping call sets which parts you need.
Yes. Your approved claims and banned terms become checks that run by code before a person reads a draft, every approval and decline is on record, and your data runs in its own environment. Suggestic's SOC 2 Type II examination is complete, with the report available under NDA, on HIPAA-compliant infrastructure.
Publish clear, well-sourced pages on your own site that answer the questions your buyers ask, and test a fixed set of those questions on the assistants before and after publishing. Do not assume the assistants will use the pages; measure whether they do. That is the work Headwater runs for you.
Profound, Semrush's AI Visibility Toolkit and Peec AI are software platforms. They do a real job of tracking how a brand shows up in AI answers, and some also draft content with an approval step. Headwater is a service Suggestic runs for you. We investigate the opportunity, write under your claims rules with checks run by code, and leave you owning what was built. If you choose Measure, we test the same buyer questions again after publishing. If you already use a platform, you can keep it. If you do not, Measure can give you that reading as part of the work.
Many shapes: articles, listicles, answer pages, product detail pages (PDPs) including content for Amazon listings, and other formats you need. We agree the formats when we set the scope. Every format goes through the same steps: sources found and graded first, code checks against your claims rules, then a person's review.
This page does not list a price. The first engagement is one question at a fixed scope and a fixed fee, and the scoping call sets the work and its cost.
A first briefing or a first measurement takes two to three weeks from the day the scope is fixed. A full build of the parts you choose takes eight to ten weeks. How soon the assistants' answers change is not something we promise; it is something we measure.
Start here
A scoping call sets the question, the sources and the deliverable. Once the scope is fixed, a first demand briefing, an assistant-answer measurement, or both takes two to three weeks and ends with one recommended next step.
Book a scoping call
