How to Filter Data by AI Model, Country & Date in AmICited

Every number in AmICited respects one control: the global filter bar at the top of the app. AI Model, Country, Tag and Date range rescope the entire product at once, so learning these filters is the key to reading every other view correctly.

Quick Steps

• The global filter bar sits in the top bar on every page: AI Model, Country, Tag, and Date range.
• AI Model isolates a single engine; Country scopes to a market; Tag scopes to a labeled group of prompts; Date range sets the time window.
• Change one filter at a time so you know exactly what moved a number.
• Combine filters (e.g. Country + Tag) to answer narrow, specific questions.
• Filters persist across pages, so check the bar first whenever a metric looks unexpectedly high, low, or flat.

What is AI search visibility segmentation?

AI search visibility is the measure of how often, and how prominently, a brand is cited by AI assistants (ChatGPT, Perplexity, Gemini, and Google AI Overviews) when people ask questions related to that brand’s category. Unlike a traditional SEO rank tracker that reports one ranking per keyword per search engine, an AI answer engine can cite zero, one, or several sources in a single response, in any order, and the answer itself can vary from one moment to the next. That variability is exactly why a single aggregate visibility number is rarely enough to act on: it tells you that something changed, not why.

Segmentation is the practice of slicing that aggregate number along the dimensions that actually explain the variance: which AI engine generated the answer, which country or language the query was asked in, which topic or campaign the prompt belongs to, and which time window the data covers. Each dimension isolates a different kind of drift. A brand can be cited constantly in ChatGPT and never in Gemini, because the two models are trained differently and weight sources differently: that’s a model effect. A brand can dominate its home market but be invisible abroad, because AI Overviews and Perplexity draw on region-specific sources and language signals: that’s a country effect. A brand’s share of voice can look strong overall but hide the fact that it’s winning informational prompts and losing commercial ones: that’s a tag, or topic, effect. And any of these can shift after a model update, a content push, or a competitor’s launch, which only a date range comparison will reveal.

This matters for generative engine optimization work because the fix for “we’re not being cited” is completely different depending on which dimension is driving the gap. If the gap is model-specific, the fix might be technical: making sure your content is crawlable and well-structured for that engine’s retrieval pipeline. If it’s country-specific, the fix might be localized content or region-specific PR and citations. If it’s tag-specific, the fix is usually content coverage: you simply haven’t published anything that answers those prompts well. Filtering isn’t a cosmetic UI feature; it’s the diagnostic step that turns a vague “our AI visibility dropped” into a specific, actionable hypothesis.

The four filters in AmICited map directly onto this framework, and because they compose (you can combine Country and Tag, or AI Model and Date range, at the same time), they let you ask genuinely narrow questions, like “how do our pricing prompts perform in the US on Perplexity over the last 30 days?” rather than only ever looking at the whole account at once.

Where to find it

The filter bar sits in the top bar, present on every page of the app: All AI Models, All countries, All tags, and a date range control (e.g. Last 30 days), positioned next to the refresh button and the + New prompt action. Because it’s fixed in the header rather than buried in a settings panel, it stays visible and adjustable no matter which view you’re in (Dashboard, Prompts, Sources, or Competitors), so you never have to re-navigate to change your scope.

What each filter does

• AI Model: scopes every chart and table to a single engine (ChatGPT, Perplexity, Gemini, or Google AI Overview) or shows all of them combined. This is the filter most likely to explain a confusing headline number: engines pull from different indexes, weight AI citations differently, and update on different cycles, so an aggregate score can mask a real per-engine problem.
• Country: scopes results to a specific market. If you track AI visibility across multiple regions, this is essential: AI answer engines localize results, and a brand that’s well cited in one country’s version of Perplexity can be effectively absent in another.
• Tag: scopes to a group of prompts you’ve labeled, whether by campaign, funnel stage, product line, or topic. Tags let you treat your prompt library less like a flat list and more like a set of named cohorts you can each report on independently.
• Date range: sets the time window every trend line, average, and delta is calculated over. This is what turns a snapshot metric into a trend, and it’s the filter you’ll reach for most often when investigating a specific change.

How to use it

1. Change one filter at a time. When you’re comparing numbers, adjust a single control so you know exactly what moved the figure. Changing AI Model and Date range simultaneously makes it impossible to tell which change caused the difference you’re seeing.
2. Diagnose per engine. Switch AI Model to each engine in turn to find where you’re strong and where you’re invisible. It’s common to see a healthy overall AI visibility score that’s really being carried by one engine while another shows almost nothing, a pattern you’d never spot looking at the “All AI Models” view alone.
3. Segment by market and theme. Combine Country and Tag to answer focused questions like “how do our pricing prompts perform in the US on Perplexity?” This kind of compound filtering is where the real diagnostic value sits: a single filter tells you that something is off, two filters together start telling you where.
4. Set the right window. Widen the Date range for long-term trend reading (is share of voice drifting up or down over a quarter) or narrow it to inspect a recent change, such as the days immediately after a content publish, a competitor launch, or a known model update.
5. Re-check the filter bar before trusting any number. Because filters persist as you move between pages, it’s easy to carry a narrow scope (say, one country, last 7 days) into a view where you meant to see the full picture. Get in the habit of glancing at the bar first whenever a metric looks unexpectedly high, low, or flat.

Reading filtered data correctly

A few practical habits keep filtered views from being misleading rather than clarifying. First, watch your sample size: narrowing by Country and Tag and AI Model simultaneously can shrink the number of underlying prompts and AI responses to the point where week-to-week swings are just noise, not signal. If a filtered chart looks erratic, widen one dimension back out before concluding anything changed. Second, treat the “All AI Models” default as a starting point, not the answer: it’s useful for a quick health check, but any real investigation into why a number moved should immediately branch into per-engine views, since engines behave enough like separate channels that averaging across them can hide the actual story. Third, when you’re building a recurring report, say, a monthly client update if you run AI visibility across an agency’s client portfolio , save the filter combination you land on as your standard view, so month-over-month comparisons are always measuring the same slice of data rather than accidentally comparing apples to oranges because a date range defaulted differently.

Tags deserve special mention because, unlike AI Model and Country, they’re something you define yourself. A well-organized prompt tracking setup usually tags prompts along at least two axes: topic (e.g. “pricing,” “comparison,” “how-to”) and funnel stage (e.g. “awareness,” “consideration,” “decision”), so that filtering by Tag can answer both “are we visible for commercial-intent questions?” and “are we visible early in the buyer journey, when brand consideration is still forming?” If your prompt library was imported or grew organically without tags, it’s worth a pass to add them retroactively; the filter bar becomes far more useful once every prompt carries at least one meaningful label. For teams still deciding how to structure that prompt set in the first place, reviewing how to find which prompts trigger AI to mention your brand is a good next step before tagging.

Where this fits in a GEO workflow

The filter bar isn’t a reporting convenience bolted onto the dashboard; it’s the mechanism that makes every other AmICited view trustworthy. A citation trend, a competitor comparison, or a source breakdown is only as useful as your confidence that you know exactly which model, market, topic, and time window produced it. Once you’re comfortable moving between these four filters, the natural next step is to look at what the filtered views are actually telling you to do, whether that’s closing a gap on a specific engine, prioritizing content for an underperforming country, or reallocating attention toward the prompt tags where your AI rank tracker shows the biggest opportunity. Master the filter bar and every other feature in this Academy becomes sharper, because you’ll always know precisely which slice of data you’re looking at before you act on it.