From AI Visibility to a Deployed Fix
Wrenda doesn't stop at telling you an AI answer left you out. It works out which page should have won it, suggests the change, serves it to crawlers with no site deploy, and checks whether the same prompts moved.
Five stages
TrackLearnOptimizeDeployVerifyTrack, learn, optimize, deploy, verify
Citation tracking and the proxy are one product, so what the tracker learns goes straight to the pages that need it, and what the proxy serves is measured by the tracker.
01
Track
Your prompts, asked of AI engines on a schedule
02
Learn
Each lost answer matched to the page that should win it
03
Optimize
Changes written for those exact questions
04
Deploy
Served by the proxy, no site deploy
05
Verify
The same prompts, before and after
Every new citation run feeds the next round of briefs. The loop does not end at a report.
Step 01 · Track
Track what AI answers say
AI Citations runs your tracked prompts on a schedule, weekly on every plan and daily on Scale, and records whether each answer names you, whether it cites your site, which competitors it names instead, and what sources it uses.
Step 02 · Learn
Turn lost answers into briefs
After each citation run, every prompt you are losing becomes a gap brief: the question, the competitors named instead, what the AI searched for and the sources it cited. Each brief is matched to the page on your site that should win it, and the pages with the most to gain are sent to the optimizer.
Step 03 · Optimize
Get changes aimed at those questions
The optimizer reads the page and its gap briefs and suggests specific changes: an FAQ that answers the lost question, an entity definition, an honest comparison, a better title or schema. Each suggestion shows which tracked questions it aims to win back.
Step 04 · Deploy
Switch it on, no release
Enable a suggestion and Wrenda serves it to crawlers from the next crawl. There is no ticket, no pull request and no site deploy. Your origin is untouched, people see your site unchanged, and switching the change off reverts it.
Step 05 · Verify
Check the same prompts
Wrenda takes a 30-day baseline for the prompts a change targets, waits for an AI crawler to re-read the page, then compares at least 7 days of new answers: won, lost or flat. The next run starts the loop again.
What each stage hands to the next
A brief for a page, a change for that brief, a verdict for that change.
Learn
Gap Briefs, Matched to Your Pages
After each citation run, Wrenda looks at the last 30 days of answers for every tracked prompt you are losing. Answers that name a competitor instead of you count most, and recommendation and comparison questions are weighted above informational ones. Each brief is then matched to the page that should win it.
- Pinned: you assign the question to a page yourself
- Cited: one of your pages is already cited for it somewhere
- Matched: Wrenda picks the page whose topic is the question
- Content gap: no page fits, so a new one is needed
- Pages with the most to gain are sent to the optimizer automatically
Also from the tracker
- Read, never cited -- pages AI crawlers fetch that no tracked answer cites: the best pages to work on
- Live fetches -- user-triggered AI fetches kept apart from training and index crawls
- Source opportunities -- third-party sites cited when a competitor is named and you are not
One lost question, one page to fix
Gap brief example
sent to the optimizerquestion: "best trail shoes for wide feet" intent: recommendation missing: 9 of 12 answers named: Northpeak (6), Ridgeline (3) searched: "wide toe box trail shoe" cited: a review site, a forum thread page: /shoes/trail-wide matched: by page topic
Sample data. Real briefs come from your own tracked prompts.
What you see in Page Optimizations
/shoes/trail-wide · FAQ · for AI crawlers
Add an FAQ: “Which trail shoes fit wide feet?”
Aims to win back
“best trail shoes for wide feet” -- missing from 9 of 12, named instead: Northpeak, Ridgeline
AI crawlers
FAQ · entity · comparison
Search bots
Title · meta · H1 · schema
Edge SEO
Canonical · OG · noindex
Optimize + Deploy
Changes You Switch On
The optimizer reads the page alongside its gap briefs: the questions, the competitors named instead, what the AI searched for and the sources it trusted. It suggests changes for that page, each one showing the questions it aims to win back. Nothing is served until you enable it, and enabling is the deploy.
- Suggestions name the tracked questions they target
- Served by the proxy to the crawlers your rules route through Wrenda
- No site deploy, no ticket, no change to your origin
- People visiting your site see it unchanged
- Disable a change to revert it
- Optional auto-apply for schema, FAQ and entity suggestions only
Verify
Did the Answers Move?
A change made from a citation gap is checked against the prompts it was made for, not your overall score. The clock only starts once an AI crawler has actually re-read the page, because no answer can reflect a version no AI crawler has seen.
- 30-day baseline of mention and citation rate for the targeted prompts
- Waits for the first AI-crawler fetch of the page after enabling
- Verdict after 7+ days and 6+ answers: won, lost or flat
- Advisory: the change stays on until you decide
- Separate Search Console check after about two weeks, if connected
Before and after, on the same prompts
How a verdict is reached
baseline: mention + citation rate for the
targeted prompts, 30 days before
the change was enabled
re-read: first AI-crawler fetch of the
page after enabling
measure: 7+ days and 6+ answers since
that fetch
won either rate up 10+ points
lost either rate down 10+ points
flat anything elseMonitoring stops at step one
Knowing you are missing from an answer is where the work starts, not where it ends.
| Monitoring-only tool | Wrenda | |
|---|---|---|
| When you are not cited | A score drops on a dashboard | The lost question becomes a brief for a named page on your site |
| Who writes the change | A brief goes to your content team | The optimizer suggests it, aimed at those exact questions |
| How it reaches AI crawlers | A ticket in the dev queue, then a site release | Served by the proxy from the next crawl, no site deploy |
| Undoing it | Another release | Switch it off |
| Did it work? | Compare charts and guess | Before and after on the prompts the change targeted |
5
Stages in one loop
30 days
Baseline before a change goes live
7 days
Minimum measuring after an AI re-read
0
Site deploys to ship a change
Closed loop, open hands
The loop does the legwork. You decide what serves.
Nothing serves until enabled
Every suggestion starts as a suggestion. Review the change and the questions it targets, then enable the ones you want, one at a time or in bulk.
- Suggested, enabled, disabled or dismissed
- Dismissed suggestions are not proposed again
- Editor or admin role to enable
Reversible by design
Changes live in the proxy, not in your codebase, so turning one off is the whole rollback. Your origin never holds a copy to clean up.
- Disable a change to revert it
- Your origin is never edited
- Shadow mode serves nothing it optimizes
Auto-apply, narrowly
Turn on auto-apply and only low-risk suggestions switch on by themselves. The elements that can hurt a page in search always wait for a person.
- Off by default
- Schema, FAQ and entity only, at 0.8+ confidence
- Never titles, meta, H1s or Edge SEO fixes
The parts of the loop
Not cited at all yet: AI search invisibility · Plans and tracked prompts: pricing
AI Visibility Loop FAQs
What changes, who sees it, and how Wrenda tells whether it worked.
No. Your origin is never edited and nothing is deployed to it. Wrenda sits in front of your site as a proxy, and an enabled change is applied to the responses crawlers get from that proxy. People browsing your site see your pages exactly as they are. Switch a change off and crawlers get the page without it again.
Not unless you turn auto-apply on, and it is off by default. When it is on, only low-risk suggestions are switched on automatically: structured data, FAQ additions and entity definitions, at a confidence of 0.8 or higher. Titles, meta descriptions, H1s, comparisons and Edge SEO fixes such as canonicals and noindex are never auto-applied. Everything else waits as a suggestion for you to review and enable.
The ones your crawler rules already send through Wrenda. Content additions (FAQs, entity definitions, comparisons) are added to the AI-optimized version served to crawlers you route to AI optimization. Title, meta description, H1 and schema changes are applied to the pre-rendered HTML served to search bots. Edge SEO fixes go into the page's HTML for crawlers. The domain has to be in active serving mode: in shadow or paused mode Wrenda serves nothing it optimizes, and the before/after check waits.
Wrenda says so. If neither a pinned page, a page AI already cites, nor its own page matching finds a page about the topic, the question is marked as a content gap. The optimizer can only improve pages that exist, so a content gap means a new page is needed. If you know which page should win, you can pin the question to it.
For changes made from citation gaps, Wrenda records the mention rate and citation rate for the exact prompts the change targets over the 30 days before you enabled it. It then waits for the first AI-crawler fetch of the page after enabling, because answers cannot change before an AI crawler has seen the new version. Once at least 7 days and at least 6 answers have passed since that fetch, the change is marked won if either rate rose by 10 points or more, lost if either fell by 10 points or more, and flat otherwise. The verdict is advisory: it is a before/after reading, not a controlled experiment, and the change stays on until you decide.
No, they are separate and can both run. If you connect Google Search Console, any enabled page optimization also gets a search check after about two weeks, comparing position, click-through rate and clicks before and after, marked won or flat. The AI-answer check described above only applies to changes made from citation gaps.
AI Citations and Page Optimizations are on every plan, including the 14-day trial. Plans differ in how many prompts you can track and how often they run: weekly on every plan, daily on Scale. For lost answers to become page changes, the brand's website needs to be one of your Wrenda domains. Enabling changes needs an editor or admin role.
Get Started
Don't just watch the answers
Track where AI leaves you out, fix the page that should win, ship it without a release, and see whether the answers moved.
Start 14-day trial14-day free trial — No credit card required
Included on every plan. See pricing