How tracking works
What a query is and how to manage your query bank, which assistants Fond asks and what you can change, when runs happen, what happens when an assistant errors, and which run counts as the latest.
Tracking is Fond asking AI assistants the questions your customers ask, on a schedule, and keeping every answer. You change what Fond asks on the Query Bank tab of the Queries page, and who Fond asks and from where on the Configuration tab. How Fond works walks through one run and what each run status means.
Manage your queries
A query is one question, written the way a shopper asks it, such as "best running shoes for flat feet". Every visibility and competitor figure covers your query bank. Fond reads each answer for your brand, how early you appear, and which sites it drew on. The Visibility guide explains the figures.
Fond writes your first queries from what you confirmed at setup, and you review them before Fond saves anything. Getting started explains each question's tier, type and focus. After setup, use the Query Bank:
- Click Add Query to add a question, with its tier, type, focus and topic cluster.
- Click Edit to change a question. A label change updates it in place. A wording change archives the old question, which keeps its history, and starts a new one.
- Click Archive to pause a question. Fond deletes nothing. Tick Show archived queries to see it, and click Restore to ask it again.
- Click Import to read a CSV file. Fond shows what each row adds or changes before you apply it. Click Export to download the bank in the same format, or a blank template if the bank is empty.
Every change applies from the next run. A past run keeps the questions it asked.
How many queries you can track
Your plan sets how many queries each website can track. The Query Bank shows your count against that limit. At the limit, Add Query is unavailable until you archive a question or upgrade.
A website can hold more queries than its plan allows, for example after a downgrade. Fond then asks the oldest queries up to the limit and marks the rest Not being asked. You can still edit those. If you archive an asked question, its slot goes to the next one. A reworded question keeps its place.
Choose which assistants are asked
The Runner Configuration card on the Configuration tab lists every assistant that Fond can ask, such as OpenAI, Gemini, Claude and Google AI Mode. You turn each one on or off, and at least one must be on.
Where the model you pick is the one that answers, you also pick the model from those in your plan.
An assistant or model outside your plan shows upgrade required. If your plan stops including something that you turned on, runs leave it out without a change to your settings.
The same card sets:
- How many times each question goes to each assistant in a run (Repetitions per query). Assistants answer differently each time, so more asks give a more stable rate.
- The market that Fond asks from (Locale). Assistants that search the web search from that country. A new website starts in the market of its language, or the United States if Fond has none. Each website has one market at a time.
Click Reset to default to return this website to the default configuration of Fond.
Fond sets how it reaches each assistant. Any change, yours or Fond's, applies from the next run. Each run records the assistants, models and market that it used.
When runs happen
The read-only Run Schedule card on the Configuration tab shows your schedule.
- On a plan with ongoing monitoring, Fond runs your queries every other day, unless Fond set another interval for your site. Runs start at about the same hour, which the card shows in your time zone with the delay a busy queue can add.
- On the free audit plan, Fond starts the one included run when you finish website setup.
Fond picks the hour at setup to spread load. Only the Fond team can change the hour and the interval.
To ask sooner, click Run Now on the Query Results page, or use the REST API. Each website can have one run in progress at a time.
Your plan caps the runs a website gets in a rolling week, scheduled and started together. A run that you start does not move the schedule. If it uses the last run of the cap, Fond skips scheduled runs until the week frees one. On the free audit plan, the cap is one run for the life of the website.
Which run is the latest
The overview and every analytics page describe the same latest run. It is the newest run that:
- finished, as Completed or Completed with errors
- has at least one answer that came back
- is not hidden
Newest means the latest run date. On the same date, the run that started later is newer. A page limited to a date range uses the latest run in that range.
The run picker on the Query Results page also lists runs in progress and failed runs, so you can see what they collected. Neither can be the latest run.
You can hide a run from the run picker, such as a test run. A hidden run leaves every chart, every period and the picker, and the app cannot bring it back.
When an assistant errors
Assistants time out and apply rate limits, so some requests fail. Fond treats a failed request as missing data, not as an answer that left you out.
- Rates that count answers use only the answers that came back.
- If one assistant failed for a whole run, its line on the per-assistant chart skips that run instead of dropping to zero. The other assistants still count.
- A run where every request failed is never your latest run, and no period counts it. You can still open it from the run picker.
The Query Results page shows which requests failed, and offers a retry where Fond can run them again. The glossary says which figures count answers and which count questions. The two handle a question whose every answer failed in different ways.
What tracking does not do
- Fond does not see what real shoppers ask. It measures only the questions in your query bank that it asks.
- Fond does not record what your customers were told. A shopper who is signed in, somewhere else, or asks on another day can get a different answer.
- Fond does not get the same answer every run. A rate over several runs tells you more than any one answer.
- Your results on an assistant that is off, or outside your plan, are not in your numbers.
Getting started
What you need before you start, what Fond reads from your site, how to review the questions it generates, what the first run does, and what to do when something goes wrong.
Visibility
What the Visibility page shows, how to read a trend drawn one point per run, why two assistants disagree about the same brand, what moves the number, and what the page cannot tell you.