Tracking & History - bing.ly Docs

A single visibility score tells you where you stand today. A series of scores over time tells you whether what you are doing is working - and when something changed that you didn't expect.

Bingly's tracking system stores every search you run against a keyword-and-domain pair, builds a timeline of visibility scores, and shows you how your AI citations have evolved. This document explains what gets tracked, how to read the history view, and how to build a tracking cadence that gives you genuinely useful data rather than noise.

What Gets Tracked

Every search you run in Bingly is automatically recorded. There is nothing to configure - the moment you run a search, that result is saved to your keyword history for that domain.

The record for each search includes:

Date and time of the search
Keyword queried
Target domain
Models tested - which AI models were included in that run
Per-model citation status - Cited / Not Cited for each model
Per-model prominence - the prominence level assigned at the time
Per-model competitor citations - which competing domains were cited
Aggregate visibility score - the final percentage for that run
Research results - if the Research feature was enabled for that run, community post counts by intent class are also recorded

This complete record means you can look back at any historical run and see exactly what the results were at that moment - not just a summary number, but the full detail.

What Is Not Tracked Automatically

Bingly does not automatically run searches on a schedule. All tracking entries are generated by searches you run manually. If you want a consistent historical record, you need to run searches consistently - the recommended cadences section below gives practical guidance on how to do this.

Reading the History View

Accessing History

The History view is accessible from the main navigation. You can view history at two levels:

All keywords - a feed of all tracked keyword-domain pairs, sorted by most recently searched
Single keyword - a detailed history view for one specific keyword-domain combination, showing the full trend chart and all historical runs

The Trend Chart

The main visual in the History view is a line chart plotting your aggregate visibility score over time. The x-axis shows dates; the y-axis shows the aggregate score as a percentage from 0 to 100.

Each data point on the chart represents one search run. Hovering over any data point shows the exact date, the aggregate score, and a breakdown of which models were cited and which were not for that run.

The trend line is more informative than any individual data point. A score of 35% after a month of content work is less meaningful than seeing that score climb from 10% to 35% over six weeks. The direction and velocity of the trend is the signal; the absolute number is the context.

The Per-Model History Table

Below the trend chart, the History view shows a per-model history table with one row per historical search run and one column per AI model. Each cell shows the citation status and prominence for that model on that date. Colour coding makes patterns immediately visible:

Green - cited (darker green = higher prominence)
Red - not cited
Grey - model was not included in that run

This table is particularly useful for spotting model-specific changes. If your aggregate score dips, the per-model table will tell you whether it was a single model that dropped out or a broad decline across all models.

Competitor Timeline

Below the per-model table, a competitor timeline section shows which domains were cited across all historical runs for this keyword. Each competitor is shown with:

The first date they appeared in your results
The last date they appeared
Their frequency of citation (percentage of runs in which they were cited)
Their average prominence

A competitor that has appeared in 90% of your tracked runs at Primary prominence is a dominant authority for that topic. A competitor that appeared once at Marginal prominence is not worth focusing on.

The competitor timeline is also where you can spot emerging competitors - new domains that have recently started appearing in AI answers for your keyword. An unknown domain that achieves consistent citation within a few months is one to study.

Setting Up Regular Tracking

The Core Principle

AI visibility changes slowly. Unlike Google rankings, which can shift dramatically overnight following a core algorithm update, AI model citations tend to shift over weeks and months as models are retrained, as new authoritative content enters the ecosystem, and as your own content changes.

Running Bingly every day for the same keyword will give you a chart with a lot of noise and very little signal. Running it every week or every two weeks will give you a chart that accurately reflects genuine changes.

Recommended Cadences

Established keywords - monthly tracking

For keywords you have been tracking for more than three months and where your score is relatively stable, a monthly tracking run is sufficient. A monthly run gives you twelve data points per year - enough to identify trends and measure the impact of content initiatives without generating noise.

Monthly tracking is right for: core product category keywords, your brand name, your main competitors' names.

Active optimisation - weekly tracking

If you are actively working on improving your AI visibility for a keyword - publishing new content, adding schema markup, building citations - weekly tracking lets you see whether your changes are having an impact within a reasonable timeframe.

Be aware that AI models typically take four to eight weeks to reflect changes in their citation behaviour after new content is published. Do not expect to publish a guide on Monday and see your score improve by Friday. Weekly tracking over a two-to-three month period will reveal whether a content initiative is working.

Weekly tracking is right for: keywords where you are actively running optimisation campaigns, new keywords you are establishing from scratch, competitive keywords where a competitor is gaining ground.

Reactive tracking

Run a search outside your regular cadence when:

You publish a significant new piece of content and want to establish a pre-publication baseline
A competitor announces a major product or content change
A search industry event (a major AI model update, a training cutoff announcement) might have changed citation behaviour
Your traffic or brand sentiment shifts unexpectedly and you want to check whether AI visibility is a factor

Reactive runs do not break your trend chart - they simply add additional data points. The date labels make it easy to annotate what was happening at each point.

Creating a Tracking Calendar

The simplest approach is a recurring monthly reminder to run Bingly for your core tracked keywords. Fifteen minutes, first Monday of the month, covers a portfolio of ten to twenty keywords at monthly cadence. If you are in active optimisation mode on specific keywords, add those to a separate weekly reminder.

A sample monthly tracking workflow:

Open Bingly and navigate to your keyword list
Run a search for each core keyword against your primary domain
Note any score changes greater than 10 percentage points
For significant drops: open the per-model table and identify which models dropped out
For significant rises: note what content or structural changes you made in the past month
Record notes in the History view (use the run notes field to annotate what was happening at the time)

Understanding Visibility Trends

Why Scores Change Over Time

AI visibility scores are not static. They change because of several factors:

Model retraining - AI models are periodically retrained on new data. A retraining event can incorporate your new content (improving your score) or can shift a model's citation preferences in ways that affect your score without any action on your part. Major model version releases (GPT-5, Gemini 2.0, Claude 4, and similar) often cause measurable shifts in citation behaviour across many domains.
Competitor activity - if a competitor publishes a definitive guide that becomes widely cited, they may displace you in model responses. Your score can drop without you having done anything wrong. The competitor timeline in your history view will show this pattern: their citation frequency will rise at the same time your score falls.
Your own content changes - publishing new content, restructuring existing pages, adding schema markup, earning press coverage or external citations - all of these can improve your AI visibility over weeks to months. Removing or significantly altering authoritative pages can reduce it.
Changes to your domain authority or entity recognition - domain migrations, brand renames, or acquiring a brand name that was previously associated with another entity can temporarily disrupt AI visibility as models reconcile the change.

Reading Score Volatility

Some volatility in your trend line is normal. Because Bingly uses multiple prompt template variants per model, and because AI model responses have some inherent variability, the same search run twice in the same day might produce slightly different scores. A fluctuation of plus or minus five percentage points between consecutive weekly runs is within normal variance.

Meaningful signals to investigate:

A drop of more than 10 percentage points between two consecutive runs
A sustained downward trend over three or more consecutive runs
A single model dropping from Cited to Not Cited and staying there
A new competitor appearing at Primary prominence for your keyword

Diagnosing a Score Drop

If your score drops significantly, use this diagnostic sequence:

Open the per-model table - is the drop isolated to one model or across multiple?
Check the competitor timeline - did any new competitors appear at the same time your score dropped?
Check your own content - were any pages changed, removed, or migrated around that date?
Check industry news - did the affected model release a new version or announce a training update?
Run the search again within a day or two - if the score recovers, it was likely a transient response variability issue rather than a structural change

If the drop is structural (confirmed across multiple runs, not explained by content changes on your end), the most likely causes are a competitor's content gains or a model retraining event. Both are addressed by the same response: improve the specificity, authority signals, and external citation of your content for that keyword.

Exporting and Reporting

Export Formats

Historical search data can be exported from the History view in two formats:

CSV export - a flat table with one row per historical search run, containing the date, keyword, domain, aggregate score, per-model scores, and top competitor citations. Suitable for import into spreadsheets, BI tools, or client reports.
JSON export - the full structured data for each run, including the complete competitor citation lists and the per-model prominence data. Suitable for integration with custom dashboards or programmatic analysis.

To export:

Open the History view for a keyword
Select the date range you want to export
Click Export

You can export individual keyword histories or a combined export of all tracked keywords.

Building a Client Report

For agency users tracking AI visibility for clients, the History view provides the data you need for a recurring AI visibility report section. A practical monthly AI visibility report includes:

Section	What to include
Score summary	Aggregate score this month vs. last month, with trend direction
Model breakdown	Which models are citing the client, which are not
Competitor landscape	Which competitors are consistently cited and at what prominence
Notable changes	Any significant movements (up or down) with context
Actions taken	Content or technical changes made this month
Recommended next actions	Highest-priority changes for next month

The CSV export maps directly to the first four sections. The last two are editorial - the data tells you what changed; your analysis explains why and what to do about it.

Data Retention

Bingly retains the full history of all search runs for as long as your account is active. There is no automatic deletion of historical data. If you close your account, you can export your full history as a JSON archive before doing so.

Connecting Tracking to Action

History data is only useful if it informs decisions. The most common mistake is running regular searches, watching the trend chart go up and down, and not changing anything as a result.

A useful mental model: treat your AI visibility score like a conversion rate. A conversion rate tells you whether your acquisition funnel is working; an AI visibility score tells you whether your content is earning the trust of AI models for the queries that matter to your business. In both cases, the number alone is not the goal - the number is an indicator of whether your underlying work is effective.

When your score is rising: identify what changed and do more of it. Which pages did you update? Which citations did you earn? What structural changes did you make? Document these because they are your playbook.
When your score is flat: check whether you are testing the right keywords. A flat score for a keyword where you have made significant content investments might indicate that the keyword phrasing doesn't match how users actually ask AI models about your topic. Try reformulating the keyword as a more specific question.
When your score is falling: treat it as a diagnostic signal, not a failure. Open the per-model table, check the competitor timeline, review the Recommendations panel from your latest run, and prioritise the highest-impact fix. AI visibility is recoverable - models update their citation behaviour as the content landscape changes, which means ground lost can be regained.