Search visibility

This is the section that answers the product's central question: does your site show up when someone asks an AI assistant about your category? AgentFi sends real queries to ChatGPT, Perplexity, Gemini and Claude, and records whether your site is cited in the answer.

How it works#

Every query is sent to all configured engines. For each query × engine pair, AgentFi stores one of three outcomes: found, not mentioned, or engine error. The sum of those outcomes is your visibility rate.

Warning

Each query costs 1 token per configured engine. Running 40 queries across 4 engines spends 160 tokens. Check the token counter in the sidebar before you press Run All.

Generating queries#

If you have none yet, the section shows its empty state with the Generate Queries button. AgentFi analyses your optimized content and writes conversational questions.

Generation options#

Query Types

Brand Queries — questions mentioning your brand name.
Discovery Queries — category-level questions, without any brand mention.

Generation Preset

Quick Scan, Standard or Deep Analysis. Each preset states how many brand and discovery queries it will generate.

Mode

Add to existing keeps your current queries and adds new ones.
Replace existing deletes the selected types and generates others.

Danger

Replace existing deletes the selected queries and all of their test results. The dialog warns you how many queries will be lost. If you want to keep the visibility history, use Add to existing.

The Advanced: select specific pages option lets you choose which pages the queries are generated from. Only optimized pages can be used; the same dialog lets you optimize them on the spot.

Custom queries#

The Add Query button lets you write a question by hand. Write it in natural language, the way a customer would ask an AI assistant:

Custom queries — examples

best project management software for startups
which agency helps with the digital nomad visa in Spain?
alternatives to [competitor] for small teams

A query has to be at least 3 characters long.

Running tests#

Button	What it does	Cost
Run All	Runs every active query on every engine.	1 token per query per engine.
Run Outdated	Only queries never tested, or tested more than 7 days ago.	The same, over fewer queries.
Run all [type] queries	One group only: brand, discovery or custom.	The same, limited to the group.
The play icon on a row	Tests a single query.	1 token per engine.
Stop All	Cancels the tests in flight.	—

Tip

Run Outdated is the sensible option for day-to-day use: it keeps the data fresh without paying again for queries you already measured this week.

Visibility cards#

Under the header, the Visibility by query type section summarizes the result in one card per type.

The large percentage

The share of test results, across all configured engines, in which your site appeared. 26 of 30 found means that of the 30 query × engine pairs measured, your site appeared in 26.

The strength label

Strong, Moderate, Needs work or No data.

The daily squares

One square per day of history and per engine. The colour reflects that day's success rate for that engine. Hover a square for the exact date and counts.

The Daily success rate legend splits the colours into five bands:

Colour	Success rate
● Red	under 25%
● Orange	25–50%
● Yellow	50–70%
● Light green	70–90%
● Green	90% or more

Reading the gap between brand and discovery

87% brand visibility next to 12% discovery visibility is a very common pattern, and it is not a bug: it means the models know you when handed your name, but do not recommend you to anyone who does not know you yet. Raising discovery is where the growth is.

Overall Visibility#

A trend chart with three aggregation modes: daily, weekly and monthly. Each point shows found / total and the matching percentage.

The query table#

Queries are grouped into three tabs — Brand, Discovery and Custom — each with its own count and visibility percentage. Within each tab there are outcome filters: All, In Progress, Positive, Negative and Favorites.

Engine columns#

Column	Engine
GPT	ChatGPT
Pplx	Perplexity
Gem	Gemini
Cla	Claude Business

The symbols in each cell:

✓ — your site was cited. The small number beside it is your position in the engine's list of sources.
✕ — not found.
— — not tested yet.

Note

The Claude column is only available on the Business plan. On the other plans a notice invites you to upgrade in order to test visibility across all four engines.

Per-query actions#

Favorite

Stars the query so you can find it quickly.

Run test

Runs this query on its own.

Cancel test

Stops a run in flight.

Delete

Deletes the query. Asks for a second confirmation. A query cannot be deleted while its test is running: cancel it first.

Query detail#

Opening a row shows the full result per engine:

AI Response

The complete text the engine returned.

Citations

The sources the engine linked, with their count.

Domain URLs found

Which of your pages specifically showed up.

Position

Your rank among the sources, and the total number of sources cited.

Type

Cited (with a link) or Text mention (no link).

Search queries used by engine

The web searches the engine ran internally to answer. This is highly useful information: it tells you which words the model is using to look for you.

Tip

Read the search queries used by engine on tests that came back negative. They usually reveal terms your content does not cover — terms you can add through custom optimization rules.

When tokens run out#

If the balance hits zero, the section shows You've used all your tokens this period next to an upgrade button. On plans with extra usage, a notice appears instead, stating the per-query cost from that point on.