A citation names you in an answer somebody else wrote, at the top of the page, with a link most people do not follow. Belper Bees sells beekeeping equipment from a unit in Derbyshire, and its page on winter feeding was named in the generated answer for how much sugar syrup to feed bees in autumn from about February 2026. Impressions held at 1,940 and clicks fell from 96 to 35.
Both halves of that are worth sitting with. Nothing was lost from the index, nothing dropped in position, and two thirds of the traffic went. The shop is more visible for that query than it has ever been and it is paid less for it.
Which queries go this way
The query that gets answered in place is the one with a single factual answer. Belper's twelve best queries split cleanly. The five that lost most of their clicks are all of the form how much, how many, when should, what temperature. The seven that held are of the form which one, is it worth, what happens if I, and where can I buy.
Query types
Why factual queries lose clicks while judgement queries hold
Belper Bees saw its twelve top queries split cleanly between single-fact lookups and decisions with local conditions.
- Five queries losing clicksSingle factual answers
- how much
- how many
- when should
- what temperature
- Sugar-to-water ratio answered inline
- Seven queries holding clicksDecisions needing context
- which one
- is it worth
- what happens if I
- where can I buy
- Syrup or fondant in a wet November
How Belper's twelve best queries split
A complete three-sentence summary satisfies a ratio lookup; conditional advice leaves readers wanting the full page.
The clicks that survive are the ones needing a decision, not a fact. A ratio of sugar to water is a fact and an answer can hand it over completely. Whether to feed syrup or fondant in a wet November in Derbyshire is a judgement with conditions in it, and an answer that tries to give it in three sentences leaves the reader wanting the page.
What being cited is worth
The 35 clicks that still arrive convert better than the 96 did. Belper's ratio of orders to visits from that query went from about one in forty to about one in eleven, so the number of orders barely moved: the people who click after reading a complete answer are the ones the answer did not satisfy, which is a narrower and more serious group.
Generated answers
What a citation gives you, against what a listing gives you
The two are not degrees of the same thing: one pays in visits and the other pays in positioning beside larger rivals.
Being cited and ranking first have separate causes, and Belper holds both a top citation and position four.
The citation also appears where a shop of this size never otherwise appears, beside two national suppliers, which is a positioning outcome and not a traffic one. Being cited and being first are separate outcomes with separate causes, and Belper is cited for the syrup query and sits fourth in the ordinary results for it.
Where this stalls: reading the fall as a penalty
The stall is a clicks graph read on its own. A page whose clicks halve while its impressions hold has almost certainly been answered in place, and a page that has been penalised or outranked loses impressions first. Count impressions against clicks, because a citation shows as an impression, and the two lines together say which of the two happened.
Reading the graph
Separating an answered query from an actual ranking drop
Reading clicks without impressions can trigger unnecessary rewrites on a page that is still visible.
- Impressions heldQuery was answered in placeCitations still count as impressions, so visibility stayed intact.
- Leave working core pages alone
- Avoid panic rewrites like Belper's March fortnight
- Support the click with judgement below the answer
- Impressions fell firstPage was penalised or outrankedLost positions cut impressions before clicks shrink.
- Inspect indexing and ranking changes
- Check competitor gains on the same terms
- Audit technical or content regressions
Belper rewrote a healthy page for a fortnight in March; the only upside was proving the citation survived.
Belper spent a fortnight in March rewriting a page that was working, on the theory that something had gone wrong with it. Nothing had. The rewrite is why the shop can now say the citation survived it, which is the only useful thing that came out of the fortnight.
Writing for the thing that quotes you
A page written as one clear answer is the kind that gets cited, and that is uncomfortable, because it means the pages most likely to be quoted are the ones most likely to lose their clicks. The practical response is to publish the number, the date and the source it applies to, because an answer needs something to attribute: Belper's page gives the ratio, the month range it applies to and the hive size it assumes, in one paragraph near the top.
Page and tracking
Structuring a page to be quoted while keeping the click
Put verifiable facts at the top for attribution, place practical judgement below, and track citations manually alongside Search Console.
- Attributed answer at topState ratio, month range, and hive size in one paragraph.Fact layer
- Conditional judgement belowGive buyers nuances a short summary cannot resolve.Click layer
- Search Console totalsTracks impressions and clicks without splitting out citations.Free since 2011
- Monthly query spreadsheetCheck a fixed list of key queries once a month by hand.2026 method
As of 2026, no standard report isolates generated-answer citations from regular listings.
A query you cannot win a click on may still be worth writing, for the same reason a sign is worth putting up. It is not worth writing as the whole of a page, and Belper now puts the factual answer at the top and the judgement underneath it, which serves both readers and gives the answer something to take.
Search Console reports the impressions and clicks and does not separate a citation from an ordinary listing, free, and it replaced the keyword data Google removed from analytics in 2011. As of 2026 there is no report anywhere that names the answers you were cited in; the free method is a spreadsheet, a fixed list of your own queries and somebody asking them once a month.
Where a page sits in the ordinary results underneath the answer is what actually decides the order, and being in the store at all is still being stored, and why a crawled page may not be. Both are inside the mechanism, before any tactic makes sense, and whether the 35 will still be 35 next year is not knowable from anything Belper can see.