Judge Keyword Viability Under AI Answers
Impressions can hold while clicks fall because the answer now sits above the results, so a plan built on volume keeps commissioning pages for queries that no longer send anyone anywhere. How much traffic is actually lost is genuinely disputed, and published figures vary widely by vertical and by who did the measuring, so the verdict has to come from your own data and a live check rather than from a headline statistic. The useful split is between queries a machine can finish and queries where the searcher still has to reach a page to act.
CATEGORY
Keyword research
FORMAT
keyword-viability-under-ai-answers.md
WHEN TO REACH FOR THIS
Use when deciding whether a query still deserves a page now that AI answers and feature blocks sit above the results.
The skill file
keyword-viability-under-ai-answers.md
---
name: keyword-viability-under-ai-answers
description: Use when deciding whether a query still deserves a page now that AI answers and feature blocks sit above the results.
---
# Judge Keyword Viability Under AI Answers
Impressions can hold while clicks fall because the answer now sits above the results, so a plan built on volume keeps commissioning pages for queries that no longer send anyone anywhere. How much traffic is actually lost is genuinely disputed, and published figures vary widely by vertical and by who did the measuring, so the verdict has to come from your own data and a live check rather than from a headline statistic. The useful split is between queries a machine can finish and queries where the searcher still has to reach a page to act.
## What you need first
- a candidate query list with the page type each would need
- live SERPs checked logged out in the target country, on mobile as well as desktop
- your own Search Console click-through history for comparable queries where an AI answer already appears
## Method
1. Check each candidate live and logged out and record three things: whether an AI answer appears, whether it cites sources, and how far the first organic result sits below it. Do this on mobile too, where the displacement is far larger.
2. Classify by answer completeness. A definition, a conversion, a date or a single fact can be finished in place; a decision that depends on price, availability, fit, personal circumstances or seeing the thing cannot.
3. For queries you already rank for, compare your click-through rate before and after the AI answer appeared on that query. Your own history is the only measurement that reflects your snippet, your vertical and your market.
4. Keep answerable queries only where being cited has value on its own, and write in the plan that the return is visibility rather than sessions, so nobody measures it as traffic six months later.
5. Prioritise queries the searcher has to complete on a site: pricing, availability, configuration, booking, and comparisons against constraints only they know. Those retain clicks because the answer cannot be finished in a paragraph.
6. Decide the measurement for each retained query before it is published - which number should move, and by when - because impressions and citations will both move without clicks following.
7. Re-check the affected queries quarterly. Feature rollout is uneven across countries and query types, and a term that was safe last quarter can lose its clicks with no ranking change at all.
## What this produces
A per-query verdict of keep, keep for citation only, or drop, with the SERP evidence and the date it was checked.
## Where this goes wrong
- Dropping a whole topic because one query in it shows an AI answer, when the commercial queries underneath are untouched.
- Using a published click-loss percentage instead of your own click-through history, when the spread across published studies is wider than the decision you are making.
- Assuming a citation brings traffic - citation frequency and click-through are separate things and need separate tracking.
---
From the QuQi skill library - https://www.quqi.io/skills/keyword-viability-under-ai-answers
Free to download · no account, no email
What you need first
-
a candidate query list with the page type each would need
-
live SERPs checked logged out in the target country, on mobile as well as desktop
-
your own Search Console click-through history for comparable queries where an AI answer already appears
Method
-
01
Check each candidate live and logged out and record three things: whether an AI answer appears, whether it cites sources, and how far the first organic result sits below it. Do this on mobile too, where the displacement is far larger.
-
02
Classify by answer completeness. A definition, a conversion, a date or a single fact can be finished in place; a decision that depends on price, availability, fit, personal circumstances or seeing the thing cannot.
-
03
For queries you already rank for, compare your click-through rate before and after the AI answer appeared on that query. Your own history is the only measurement that reflects your snippet, your vertical and your market.
-
04
Keep answerable queries only where being cited has value on its own, and write in the plan that the return is visibility rather than sessions, so nobody measures it as traffic six months later.
-
05
Prioritise queries the searcher has to complete on a site: pricing, availability, configuration, booking, and comparisons against constraints only they know. Those retain clicks because the answer cannot be finished in a paragraph.
-
06
Decide the measurement for each retained query before it is published - which number should move, and by when - because impressions and citations will both move without clicks following.
-
07
Re-check the affected queries quarterly. Feature rollout is uneven across countries and query types, and a term that was safe last quarter can lose its clicks with no ranking change at all.
What this produces
A per-query verdict of keep, keep for citation only, or drop, with the SERP evidence and the date it was checked.
Where this goes wrong
-
Dropping a whole topic because one query in it shows an AI answer, when the commercial queries underneath are untouched.
-
Using a published click-loss percentage instead of your own click-through history, when the spread across published studies is wider than the decision you are making.
-
Assuming a citation brings traffic - citation frequency and click-through are separate things and need separate tracking.
Use this skill in your own AI
The download is a plain markdown file with the name and trigger in its frontmatter. Where an assistant supports skills it can load itself, that frontmatter is what it reads to decide this one applies.
Claude Code
Save it as ~/.claude/skills/keyword-viability-under-ai-answers/SKILL.md and Claude loads it on its own when what you are doing matches the trigger line. Put it in .claude/skills inside a project instead if the whole team should have it.
Claude
Upload the file in the skills section of your settings. Once it is there it applies itself in any conversation where the trigger fits, so you do not have to remember it exists.
ChatGPT
There is no skills format to install into, so paste the file contents into a Project instruction or a Custom GPT instead. It then applies to every chat in that project rather than only the one you paste it into.
Anything else
Paste the markdown into the chat before your question. It works in any assistant, it just has to be pasted again each time.
Questions about this skill
When should I re-judge a query rather than plan from its volume and impressions?
When an AI answer or a large feature block sits above the results. Impressions can hold while clicks fall, so a plan built on volume keeps commissioning pages for queries that send nobody anywhere. How much traffic is lost is genuinely disputed, and published figures vary widely by vertical and by who measured, so the verdict has to come from your own data.
What do I need in hand before starting?
A candidate list with the page type each query would need, live results checked logged out in the target country on mobile as well as desktop, and your own click-through history for comparable queries where an AI answer already appears. Check desktop alone and you understate the displacement badly, since the first organic result sits much further down on a phone.
What do I end up with, and which part gets used?
A keep, keep for citation only, or drop verdict per query, with the SERP evidence and the date checked. The citation-only category is the one that has to be written down, because the return there is visibility rather than sessions and somebody will otherwise measure it as traffic six months later. The date matters as well, since feature rollout is uneven and verdicts expire.
What is the mistake that ruins this, and what does it cost?
Dropping a whole topic because one query in it shows an AI answer. The definitional query at the top can be finished in place while the pricing, availability and configuration queries underneath are untouched and still convert, so you delete the commercial half of a cluster. The other costly assumption is that a citation brings traffic; citation and click-through need separate tracking.
More in Keyword research