Every drop comes with a reason
Other tools show you the dip. Askority tells you which engine moved, which prompts you lost, who took the slot, what changed on their side, and whether the move is real or noise. Then it ties AI visibility to sessions and conversions in GA4.
Why is a chart not enough?
A visibility line that goes down at 9am and up at 5pm teaches you nothing. The value is in knowing what to do about it, and that needs a cause, a confidence level and proof that the work pays off.
A dip without a cause is a guess
You see -11 points. Was it one engine? One competitor? One page they published? Without the diagnosis you spend a week finding out what a tool should tell you in a sentence.
Half of the moves are noise
AI answers vary run to run. If a tool samples once a day, a bad roll looks like a trend. You end up reacting to nothing, or ignoring a real loss.
Proof the work mattered
Your CFO does not care about share of voice. They care about sessions and signups. Attribution to exact landing URLs closes that loop.
Click the red dot. Get the diagnosis, not a support ticket
Every significant move on the visibility chart gets a marker. Open it and Askority lists the engine deltas, the prompts you lost, the competitor now in your slot, and the likely root cause based on what changed in the cited sources.
- Engine deltas separate a real Perplexity drop from ChatGPT noise.
- Prompts lost with who took the slot so you know exactly what to win back.
- Root cause points at the page, review or thread that the engine started citing.
- Direct link to the fix opens the matching action in your queue.
Numbers that admit how sure they are
Each score carries the number of runs it is built on. A 15-point swing over 24 runs is a signal. The same swing over 2 runs is a bad sample, and the interface says so instead of drawing a dramatic line.
- Run count shown next to every number, on every screen.
- Low-sample flags stop you from acting on a single odd answer.
- Stable vs real drop labels are computed, not eyeballed.
| Day | Runs | Visibility | Read as |
|---|---|---|---|
| Mon | 24 | 81% | stable |
| Tue | 24 | 79% | stable |
| Wed | 2 | 64% | low sample |
| Thu | 24 | 80% | stable |
| Fri | 24 | 67% | real drop |
From AI answer to session to conversion, by exact URL
Connect GA4 and Askority attributes sessions and conversions coming from AI engines by engine, by region and by the exact landing page. Not a modelled estimate: the real referral data, organized around the same prompts and engines you track.
- Three KPI cards sessions, conversions and conversion rate from AI, with deltas.
- By engine and by region tables to see which market the AI traffic comes from.
- Landing pages with exact URLs so you know which page the engines send people to.
| Engine | Landing URL | Sessions | Conv. |
|---|---|---|---|
 chatgpt.com |
/pricing |
1204 | 31 |
 perplexity.ai |
/integrations |
688 | 14 |
 copilot.microsoft.com |
/blog/crm-for-small-business |
402 | 9 |
 gemini.google.com |
/compare/acme-alternative |
311 | 7 |
Every answer stored. Every number auditable
Behind each score is a log of executions: the date, engine, position, whether you were mentioned, and the answer snippet with your brand marked. Anyone on the team can open the log and read the actual answer that produced the number.
- Executions log per prompt and per page, with the full answer on request.
- Brand marks highlight you and competitors inside the text.
- Full analysis breaks a single answer into mentions, positions and cited sources.
| Run | Engine | Position | Result |
|---|---|---|---|
| Feb 27 10:12 | ChatGPT |
#1 | Mentioned |
| Feb 27 09:40 | Perplexity |
- | Not mentioned |
| Feb 27 08:05 |  Claude |
#2 | Mentioned |
| Feb 26 22:31 | Gemini |
#3 | Mentioned |
Questions we get asked
How do you find the root cause of a drop?
We compare the sources engines cited before and after the move. When a new page starts appearing across the lost prompts, that page is the likely cause. We show it with the prompts it affected.
What counts as a significant drop?
A change larger than the normal run-to-run variance for that prompt set, measured over enough runs to be confident. Small daily wobble is not marked.
What if the drop is on my side, not a competitor's?
Diagnostics also flag your own changes: a page that stopped being cited, a bot that started getting 403s, a pricing page that went behind a login.
How does GA4 attribution work?
Read-only connection to your GA4 property. We classify sessions by referrer and landing page and group them by AI engine, country and URL. Nothing is written to your GA4.
Why show the run count at all?
Because two tools can show the same score with very different reliability. We would rather show you an honest 2-run number flagged as low sample than a smooth line.
Can I export the evidence?
Yes. Executions, prompts and attribution tables export to CSV. Full answers are available per run.
Stop guessing why the line moved
Connect a brand and GA4, and see the first diagnosis on your next drop.
Start for free