The product
What Bungad does
Bungad asks the questions your buyers ask AI, records whether the answer named you, works out who it named instead, and writes the fixes. This page is the whole product, feature by feature, including the parts that are not built yet.
The loop you are buying
Everything below belongs to one cycle, run again every month. The point of the cycle is that the last step changes the first: the fixes you publish are what the next run measures.
- Know the questions your buyers ask about your category.
- Ask them of the answer engines.
- Measure whether you appeared, where, and whose pages were cited.
- Diagnose why not, and who won the answer instead.
- Fix the specific thing that caused it.
- Prove it moved, on the next run.
Your questions, written the way buyers ask them
Setup takes one sentence: what you sell. From it Bungad drafts your starting questions, typed by what the buyer is doing — discovery, price, comparison, reputation, getting started — because those produce different answers and you want to be in all of them.
They are a starting point, not a verdict. Edit any of them, add your own, and switch off the ones that stop mattering. One rule is enforced rather than suggested: a question may not name your own brand. An answer that mentions you because the question did measures nothing, and a tracker that lets you do it is selling you a flattering number.
What gets measured in an answer
Every tracked question is asked of every engine we hold a key for, and each answer is read for four things.
- Visibility score. One composite number, weighted: how often answers name you, how often the sources they cite are yours, and where you rank when you are named.
- Answers naming you. The share of scored answers that mention your brand at all.
- Citation share. Of every source the engines cited, the share pointing at your own site — always shown as a count over its denominator, never as a bare percentage, because a percentage with no denominator hides whether anything was cited at all.
- Average position. Where you rank among the brands an answer names, because being mentioned last is not the same as being the recommendation.
Underneath those, each answer is stored in full with every source it cited, the first place your brand appears in the text, and the sentiment of the sentence it appears in. Being named badly is not the same as being named well, and the stored answer is what lets you check our reading against the engine's own words.
Who won the answer instead of you
Name the brands you lose deals to and Bungad looks for them in the same answers. You get share of voice across your tracked questions — how often each rival is named beside you — and, per question, the domain that actually supplied the answer.
That last part is the useful one. An answer engine does not invent its recommendation; it reads a page. Knowing it read a review site, a directory or a competitor's pricing page tells you what to publish.
The fixes, with the evidence attached
Bungad turns each run into a list of specific actions, ordered by impact and effort, each quoting the run and the answer it came from. The rules behind them:
- Win the answer for a question no answer named you in, naming the domain that won it and what to publish.
- Close the gap on a named competitor for a specific question.
- Become the cited source when you are mentioned but your own pages are not quoted.
- Move up the answer when you are named after your rivals.
- Get listed on a directory or review site the engines keep quoting back.
Mark one published or decided against, and the next run measures what changed. An action with no evidence under it is an opinion, so every one of them carries the question and the answer that produced it.
Whether engines can read your site at all
Before any of the above matters, a crawler has to be able to fetch your
pages. Bungad reads your robots.txt, your
llms.txt and your homepage, and tells you which answer
engines are allowed in.
This is the only part of the product that is certain. Everything
else is inferred from what an engine said on one day. A rule in your own
robots.txt blocking a crawler is the fact itself, and
nothing else on your report can improve while it stands. It also catches
the quieter version: a page that is empty until JavaScript runs, which
most answer-engine crawlers never execute.
Reports, history and getting the data out
Every run is kept. The report shows what moved since the last one and which questions moved it, and it prints to PDF for clients. Run history and individual runs export to CSV, so the numbers are yours rather than ours. Checks run monthly on a schedule on paid plans, and on demand whenever you want one.
What we do not do yet
Published here rather than left for you to discover.
- Automated tracking covers the engines we hold an API key for. The roster, and which of them are keyed today, is printed at the foot of every screen in the app and on our coverage page. Where an engine is not keyed, nothing about it is measured and nothing about it is claimed.
- Google AI Overviews and Microsoft Copilot have no public API. They are reviewed by hand for Scale clients, and that is a person looking, not software measuring.
- No alerts yet. You find out what changed when you open the app or read the monthly report, not the moment it happens.
- No team accounts. One login per account today.
What comes on which plan
The limits that differ between plans are the number of questions you track, the number of brands, how many checks a month you can run on demand, and whether every keyed engine is queried or one. The pricing page lists them against each plan. Every feature described above is on every paid plan; what changes is how much of it you get.