Prompt history
Track how your brand appears in ChatGPT and AI search over time
Rankry keeps every run of every prompt you track. Open a prompt and see, run by run and engine by engine, whether AI named you, where it placed you, who it chose instead, and what the answer said.
7-day Base trial · No card needed
Best app to track UGC brand deals?
Last 90 days7/ 12 runs
Runs that named UGC Planner
- ChatGPT · latest runMentioned · #2
- Perplexity · latest runMentioned · #2
- Google AI Mode · latest runMentioned · #1
- AI Overviews · latest runNot mentioned
Sample data
In short
Prompt history is the record of every time a tracked prompt was asked to each AI engine. Rankry keeps every run for as long as your plan’s history allows (90 days by default, up to 12 months on plans that include it), so you can see when an engine started or stopped naming you, which brand took your place, and what the answer actually said.
Inside the prompt panel
What does prompt history show for each prompt?
Click any prompt on the Prompts page and a side panel opens with its full history. Everything in it comes from answers Rankry already stored, so nothing has to re-run when you look.
A header and four tiles
The header sums up the history, for example “12 daily runs · Oct 8 – Oct 19 · 5 models”. Four tiles follow: “Runs that named you”, “Visibility” averaged over the period, “Position when named” and “Models that named you”.
Who wins this prompt
A table of every brand the answers named, with its “Share of answers” and its average position “When named”. One click adds a brand to your competitors or removes it.
Your place in each model’s answer
A run-by-run grid with one row per engine. Each cell shows whether that engine named you and at which rank. Long histories fold into weeks or months so the pattern stays readable.
A run history table
One row per run, with “Named by”, “Position when named”, “Visibility”, “Brands” and “Sources”. A quarter of results fits on one screen.
Every engine’s answer
Open a run to see each engine’s result: “Mentioned · #2” or “Not mentioned”, how many sources and brands it used, a sentiment chip, and the brands in the order the model ranked them, each with its reason.
The full answer
For every new run, the panel also keeps the engine’s complete reply under “Full answer”, so you can read how it described you and your competitors in its own words.
Set it up once
How do I track how my brand appears in ChatGPT over time?
You don’t need to save screenshots or ask the same questions by hand every week. Set up tracking once and the history builds itself, for ChatGPT and every other engine your plan includes.
- 01
Add the questions buyers ask
Rankry writes a starting set of prompts from your website at onboarding. Add your own in the “Add prompts” window or import a CSV. New prompts are answered right away.
- 02
Let the schedule run
Your project re-runs automatically every day, or every 3 or 7 days if you choose. Each run asks every prompt once on each engine it is set to use, and stores the result.
- 03
Pick a period
Choose “Last 7 days”, “Last 30 days”, “Last 90 days”, “Last 12 months” or a custom range. The period is trimmed to the history your plan keeps.
- 04
Open a prompt and read the trend
The grid shows where each engine started or stopped naming you. Open the run where something changed to see who replaced you, which sources the engine used and what it wrote.
What gets stored
What does Rankry keep for every run?
A run is one pass over your prompts. For each prompt and engine, Rankry stores the items below. A few were added recently, so older runs may not have them.
| Item | What you see | Kept for |
|---|---|---|
| Mention and rank | “Mentioned · #3” or “Not mentioned” for each engine | Every run |
| Brands in the answer | Each brand in the order the model ranked it, with the reason it gave and the sources tied to it | Every run |
| Sources | The pages the engine cited or consulted, when it searched the web | Every run |
| Sentiment chip | Positive, Neutral or Negative, from labeled claims the answer makes about your brand | Runs since Oct 13, 2026 |
| Full answer from Google surfaces | The complete AI Overviews or AI Mode answer, with the capture date, location, language and device | AI Overviews since Oct 4, AI Mode since Oct 8, 2026 |
| Full answer from chat engines | The complete reply from ChatGPT, Claude, Gemini, Perplexity or Grok | Runs since Oct 22, 2026 |
| Full answer from Meta Muse and Copilot | The complete reply, as the engine returned it | Every run |
Older runs show “Not stored for this run” in place of the full answer, and no sentiment chip. Their brands, ranks, reasons and sources are kept.
Diagnose a change
Why did my AI visibility drop?
Most drops trace back to a few causes, and prompt history lets you tell them apart. Start by checking whether the drop is broad or narrow. If one engine stopped naming you across many prompts at once, the cause is usually on that engine’s side: a model update, or a change in the sources its web search returns. If several engines dropped you on a handful of prompts, the cause is more likely in that topic.
Next, open “Who wins this prompt” for the prompts you lost. When the same competitor shows up in the runs where you disappeared, it has usually gained new coverage, such as a review roundup, a comparison page or a forum thread that the engines now cite. Open the sources of that run to find the page.
Then read the full answer. Sometimes you are still in it, only lower down, or described with an outdated detail. Sometimes the engine answered without searching the web and fell back on what it already knew. Each case calls for a different fix, from updating your own page to earning a mention on a source the engines trust.
Keep single-run swings in proportion. Each run asks each engine once, and AI answers vary from one ask to the next. A change that holds for two or three runs, or that shows up across several prompts, is a real shift. One missing mention on one engine usually is not.
History length
How far back does AI visibility history go?
History depends on your plan. These are the settings that decide what you can look back on.
| Setting | How it works |
|---|---|
| History window | Set by your plan: 90 days by default, and up to 12 months on plans that include it. |
| Period presets | “Last 7 days”, “Last 30 days”, “Last 90 days”, “Last 12 months”, or a custom range. Longer ranges are trimmed to your plan’s window. |
| Runs in the prompt panel | The newest 120 runs of a prompt, within your plan’s history window. At one run a day, 120 runs span about four months; a 3-day or 7-day interval spreads those runs over a longer period. |
| Dashboard metrics | Visibility, Position, Sentiment, Share of voice and Diversity are computed from every answer in the period you pick. Each headline metric has its own history view. |
| Run schedule | Daily by default, with the option to refresh every 3 or 7 days. New prompts get their first answers as soon as you add them. |
Before and after
Can I compare AI answers before and after a content update?
Yes. Because every run is kept, the runs before your change are your baseline. Rankry doesn’t mark your own publishing dates on the charts, so keep a note of them and follow this routine.
- Write down the date you publish or update the page.
- Pick a period that covers several runs before and after that date.
- Open the prompts the page was written for and compare the run grid on both sides of the date.
- Check whether your page now appears among the sources of the newer runs.
- Read the full answers to see if the wording about you changed, not only your rank.
- Give it time. Engines pick up new pages at different speeds, and one run is not a trend.
Explore Rankry
Related features
Does Rankry keep every run of a prompt?
Yes. Every run is stored for each engine the prompt runs on, including the first answers a new prompt gets when you add it. You can browse them for as long as your plan’s history allows, which is 90 days by default and up to 12 months on plans that include it.
How much do AI answers change from one run to the next?
There is no fixed amount. Answers built from a web search change when the search results change, and every model varies its wording and its picks between asks. Some prompts stay stable for months; others move every run. That is why Rankry shows trends across runs and engines instead of judging you on a single answer.
Can I get alerts when my AI visibility changes?
There are no separate change alerts today. On paid plans Rankry emails you when each new report is ready, and the dashboard and prompt history show what moved. Open the prompts that changed to see which engine dropped or added you and why.
Why do older runs say “Not stored for this run”?
Rankry started keeping the full reply word for word on October 4, 2026 for AI Overviews, from the first AI Mode run on October 8, 2026, and on October 22, 2026 for ChatGPT, Claude, Gemini, Perplexity and Grok. Runs before those dates still keep the brands each engine named, their order, the reasons given and the sources cited, but not the reply itself. Meta Muse and Copilot keep the full reply from their first run.
Can I see history for a competitor?
Yes, for any brand the answers named. “Who wins this prompt” shows each brand’s share of answers and average position over the period, and the Compete page shows trends for you and your tracked competitors. When you add a competitor, its figures are worked out from stored answers instantly, as far back as your history goes.
Does looking at prompt history use requests?
No. History reads answers that are already stored. Requests are spent only when prompts are asked: one request for each prompt on each engine in each run. Failed answers don’t use requests either.
Can I edit a prompt without losing its history?
Prompt text can be edited while you review your prompts during onboarding. After that, prompts are added and removed rather than edited, which keeps each history consistent: every run in it answered the same question. To track a new wording, add it as a new prompt and it is answered right away.
Is prompt history the same as the Sources page?
No. Prompt history covers every run of a prompt within your period. The Sources page analyzes the domains cited in your current report only, with no date range. To see which sources an older run used, open that run in the prompt panel.
Can I export my AI visibility history?
The project’s “Export” menu downloads your report data as CSV, packed in a ZIP with visibility and position in one file and sentiment in another. For a quick look, the prompt panel’s run history table is often enough.
See how AI answers about you change, run by run
Add your site and get your first report in minutes. Every run after that is kept.
7-day Base trial · No card needed