In 2026, the first vendor shortlist a software buyer sees is often written by a model. Search is shifting from ranked links to answers written by language models, and a brand that is missing from those answers is missing from the shortlist.
That makes AI search a channel you have to measure, the practice known as generative engine optimization. To track your brand there, you ask the questions your buyers ask, on a schedule and the way a buyer sees them, then count which companies each answer names and links. This guide covers what to measure, how to set up the monitor, and what it needs to run every week.
Why Does AI Search Visibility Matter Now?
Buyers have moved their research. A growing share of software buyers now start their research in an AI chatbot instead of a search engine, and many ask it to build the shortlist for them.
The old way to check visibility was a rank tracker: one keyword, one results page, one position. An AI answer has no fixed position. It names a handful of vendors, links some of them, and changes by engine, by phrasing and by week, so the only reliable way to know what it says is to ask it yourself.
That changes what your team measures. A brand is either mentioned in the answer or not, and either cited with a link to its site or not. Both numbers move from week to week, and both improve with content once you can see where you are missing.
What Should a Brand Monitor Track?
A brand monitor turns a pile of answers into a few numbers your team can act on every week. The figure below lists what Pinetree Research's monitor counts and where each count goes.
Answer share. The share of answers that name your company at all. It is the headline number, because a buyer cannot shortlist a vendor the answer never mentions.
Citation rate. The share of answers that link your site. It is stricter than a mention, and the link is what sends a reader on to your pages.
Gaps. Every question where competitors are named and you are not. Each gap becomes a brief for an article that answers that question directly, which makes it the most useful output of the whole monitor.
Top cited domains. The sites the engines link most often across your questions. They show you where a mention, a guest piece or a listing would lift your citation rate.
The weekly trend. Each run lands in a trend table next to last week's. You read a change against the previous run rather than memory, so you can tell a real shift from a one-off answer.
How Do You Set Up an AI Search Monitor?
You set one up in four steps: write the questions, ask each engine the way a buyer does, keep evidence of every answer, and score the results on a schedule.
Write the Questions Buyers Ask
A monitor is only as good as its questions, so start there rather than with tooling. Pinetree's question set comes from Solari's keyword families: category terms such as "best cloud browser infrastructure for AI agents", superlatives, use cases and comparisons, phrased the way a platform engineer would type them. Each question carries the competitors that could plausibly appear in the answer.
If you sell into more than one market, score each market as its own ranking. Solari sells into browsers and sandboxes, so its set is scored as two separate rankings, each against its own vendors, and a strong showing in one market never hides a weak one in the other.
Ask Each Engine the Way a Buyer Does
Many buyers ask a chat assistant in a browser tab without signing in, so that is the answer to check. Pinetree's monitor asks ChatGPT logged out through a real headless browser, and asks Gemini, Claude and Perplexity over their APIs with web search turned on.
Start every question from a clean state. A browser that carries saved cookies or history from earlier runs can bend the answer, so each question opens a fresh session and closes it once the answer is read.
Keep Evidence of Every Answer
Answers drift, so a score without the answer behind it is hard to explain. Record which companies each answer names and which sites it links, and take a screenshot of the page as a quick visual record.
Score and Act Every Week
Run the full set on a schedule, score it, and send every gap to whoever writes your content. The monitor pays off when this week's gaps become next week's articles.
What Role Does Solari Play?
Solari is the infrastructure the monitor runs on: cloud browsers, sandboxes and desktops from one API. Pinetree Research, the company behind Solari, runs its own AI search monitor this way, inside its marketing system on Solari.
A clean browser for every question. A Solari session is a real Chromium browser you drive with the standard Playwright API. Every option is opt-in, so a default session starts with no profile, recording, proxy or stealth, and each question begins with nothing saved to bend the answer.
Sessions that start at once. Solari browsers are ready in 8 ms, so a fresh session per question adds no wait to a long question set. Your monitor closes each one once the answer is read and opens the next.
Answers from your buyers' country. Route a session through a proxy in the country you choose, so you see the answer a buyer in that market sees. Turn on stealth with it, and the session confirms the country and timezone it was given.
A replay of every answer. Record a session by setting one flag at launch, then replay it in the console or download it as NDJSON, a line-by-line log of everything on the page. Recording captures input values by default, which is harmless here because the only thing typed is a public question.
A sandbox that runs the schedule. A Solari sandbox is an isolated Linux machine you create by API, and by default each one is a microVM of its own, never shared with another organization. A full pass starts as a background execution, so its output and exit status survive your client disconnecting, and an idle sandbox pauses between runs instead of being killed.
Browsers, sandboxes and desktops run from one Solari account and one bill, so the monitor needs no second vendor. Solari is also about 2x cheaper than Browserbase per browser hour, so a weekly question set is cheap enough to leave running. A companion piece on how Pinetree runs its full marketing system on Solari covers the other agents.
Where Is AI Search Tracking Going Next?
Today most teams check AI answers by hand, a few questions at a time. Over the next year that changes in four ways:
From weekly to daily. As engines update more often, monitors will run every day and flag a drop the morning it happens.
From reports to drafts. A gap will go straight to a writing agent that drafts the answer article, with a person approving it before it ships.
More places to ask. Assistants are moving into browsers, phones and operating systems, and each new surface becomes another engine to track.
Agents as buyers. Agents will build vendor shortlists on a person's behalf from the same sources the engines cite, so being cited will matter even when no person reads the answer.
What stays human is judgment: which questions matter to your buyers, what your company claims, and whether an article is true before it goes out. The monitor shows where you are missing; people decide what to say.
Get Started With Solari
Start with the questions a buyer would type about your category, run them through a Solari browser from a sandbox once a week, and keep the companies and links each answer names. Solari is the fastest and cheapest way to run cloud browsers and VMs for brand-monitoring agents. Get started with a free account and start running your agents today.
