Home / Answer log
Free tool updated 2026-11-22

Track what AI says about you, by hand.

Ask your buyer questions in ChatGPT, Perplexity, Gemini or anywhere else. Paste each answer into the Answer check with your brand and domain, and save it here. This page shows how often answers name you, recommend you and cite your own pages, over time.

How it counts:

Track the questions that stand for real buying decisions, not dozens of variations of each: two wordings of one decision ("best CRM for startups" and "top CRM tool for startups?") count as two questions, so that decision weighs twice. The page points out wordings that look like one decision, and discovery questions that read like a how-to or a definition ("how to close deals faster") and may not stand for a buying decision. Both are word matches, so check them yourself. It also lists the questions where a competitor is named in more answers than you, widest gap first, each linked to the question doctor.

The answers stay in this browser: export them to keep a copy, or to use earncitations track report on the same file. "Copy a summary" puts the same report as plain text on your clipboard, with what's still to ask, for a team update.

Check and save an answer

What to ask next

Load the question set you froze on the question map ("Download as a question set"), or write your own below. For each engine you ask, this lists the questions with no answer saved in the last few days. Each one opens the Answer check with the question and engine filled in, and answers saved for a question in the set keep its id, so a reworded question never passes for the old one.

When an engine doesn't answer (it refuses, errors, is unavailable, or asks a question back), choose why under "Didn't answer?". The attempt is saved, the question stays on the list, and the report above shows how many attempts got an answer next to every rate. A rate over the few answers that came back can look good while most attempts failed.

Above the list, a line says what your routine gives you: about how many answers a month (questions × engines × how often you ask), how soon the before-and-after compare can read a question, and how far a rate can move between two months by chance. The swing is the usual 95% margin for the difference between two rates near 50%, as if every answer were independent. Repeats of one question aren't, so read it as the least a rate wobbles, not the most.

No question set? Write your own

One question a line, in your buyers' words: what they'd type before choosing anyone. A few questions that stand for real buying decisions beat dozens of variations. Questions that name your brand or domain are marked diagnostic, the rest discovery. To see your rates per buying decision, put a line like ## Pricing above the questions about it; any other line starting with # is a comment. At most 50; the set is frozen today and kept in this browser.

Before and after a change

You shipped a fix (a pricing FAQ, a corrected page) and want to know whether answers moved. Give the day it went live. For each question on each engine, this compares the answers saved before with the ones after: improved, unchanged, worse or inconclusive. It needs at least three answers on two or more days on each side, because answers vary from run to run and repeats on one day aren't independent. A reworded question is a new question, so it's never compared with its old wording. It shows a sequence, not proof: first check the change was delivered with the delivery check.

Good practice