Local SEO · Answer-shaped content at scale
When the clicks fall and the AI citations quadruple
A city guide with 3.9 million Google impressions, no advertising budget, and one page that pulls 396,968 impressions at a 0.09% click rate while being the most-quoted page on the whole site. This is the clearest evidence I have that page format decides which channel you win — and that a clicks report can no longer tell you whether a page is working.
3.9M
Google search impressions in 16 months
38,249
Organic clicks from those impressions
60,500
Appearances in Google AI features
40,101
Bing Copilot citations in 180 days
35
Distinct pages cited by Copilot on an average day
Property
CoimbatoreJunction.in · owned and operated
Type
Hyperlocal city guide, 8 topic pillars
Live since
2020, published continuously
Window
16 months, ending 12 August 2026
Paid promotion
None. No ads, no bought links, no placements
Written by
Residents of the city, for first-hand accuracy
The finding
The page that looks broken is the most-quoted page on the site
Most case studies open with a number going up. This one opens with a number that went down, because that is where the useful part is.
Organic clicks on this property peaked in mid-2025 around a one-off demand event — a major concert in the city — and then settled to a lower baseline. Nothing unusual: event demand arrives, then leaves. But underneath that, something else was happening. Over the same period the site's reference pages were increasingly used to answer queries rather than to receive visits.
Google generative-AI appearances reached 60,500, with daily volume roughly quadrupling across the reporting window. Bing Copilot citations hit 40,101 in 180 days — averaging 222.8 a day across 35 distinct pages — and rose 45% from March to July 2026.
One page, one impression pool, two very different outcomes
/explore/red-taxi/ — 47-question reference guide with pricing tables. Everything below comes from the same
396,968 Google impressions.
Impressions
396,968
Times this page appeared in Google search results over 16 months — about 10% of the entire site's impressions, from a single URL.
What a rankings report shows you. On this number alone the page reads as broken, and the usual recommendation would be to rewrite the title or cut the page.
That is 60% of every AI appearance the site earned, from one URL. The answer was assembled from this page and delivered on the results screen, so the visit never needed to happen. Same page, same demand, different delivery.
Why one page wins clicks and another wins citations
Two pages, near-identical impressions, opposite outcomes
This is the control experiment the property accidentally ran. Both pages sit around 400,000 Google impressions. Their formats differ, and so does the channel each one wins.
| Page | Format | Impressions | Clicks | CTR | AI appearances |
|---|---|---|---|---|---|
| /explore/red-taxi/ | 47-question reference guide with pricing tables | 396,968 | 362 | 0.09% | 36,386 |
| /coimbatore-ward-list/ | Short lookup table people need to open and read | 401,847 | 5,714 | 1.42% | 11,529 |
The ward list is a lookup. You need to find your own ward in a long table, so the answer cannot be summarised for you — you have to open the page. It earns 5,714 clicks at a 1.42% CTR and 11,529 AI appearances. It wins both.
The red taxi guide is a reference: 47 questions with real pricing tables covering per-kilometre rates, hourly packages and outstation routes. Every one of those is a single-fact question — what does it cost per km? — and a single fact is exactly what an AI Overview can lift and display. So it does, 36,386 times, and the click stops being necessary.
The lesson is not "write shorter pages". It is that question-shaped, fact-dense content is citation-optimised and click-hostile at the same time, and you have to decide which you were building for before you judge the result.
The constraints
Competing with the government for its own information
A city guide has no structural advantage. It has to earn every position against the organisation that issued the fact in the first place.
The incumbents are institutions
A city guide competes with the municipal corporation, the transport authority and the local newspapers — organisations with decades of domain authority and the official answer. You are not going to out-authority them, so you have to out-answer them.
Every fact has an expiry date
Bus timings change, hospital packages get repriced, event dates pass. A page that was correct in March is wrong by September, and a wrong local answer is worse than no answer. Freshness had to be a maintenance schedule, not an intention.
Zero paid amplification
No ads, no sponsored placements, no link buying. Which is a constraint and also the whole point: it isolates content structure and maintenance as the only variables, so the result is attributable to them.
What was actually built
Six rules, none of which cost money
Nothing here required a budget. It required deciding what to write from what people search for, building the page around the answer, and keeping it correct afterwards.
Pages come from queries, not a calendar
Every page exists because people were already searching for that specific thing — ward numbers, a hospital package price, a bus route, a snake catcher. Nothing was commissioned because it seemed like a good topic.
The answer goes above the fold
The number, the table or the list comes first, in scannable form. No 400-word preamble before the bus timings. This is the decision that produced both the click-through rates and the citations.
Real tables, never images
Prices, routes and timings are marked-up HTML tables. An image of a table is invisible to a crawler and useless to a model. This one technical choice is why the pricing pages get quoted verbatim.
Written by people who live there
First-hand accuracy on things you cannot verify remotely: which counter to go to, what the office actually opens at, whether the bus still runs. That is the part an aggregator cannot copy.
Updated before demand peaks, not after
Event pages, exam dates and seasonal fees are refreshed ahead of the search spike. Freshness treated as a product feature with a schedule attached.
Sequential searches linked together
Someone who looks up a bus route needs the timings next, then the fare. Internal links follow the actual sequence of a task rather than a topical taxonomy.
Evidence · Search Console
38,249 clicks from 3,913,641 impressions
Note the shape of the first chart: the mid-2025 peak is the concert, and the decline after it is the event demand leaving. That is the trend the AI numbers later contradict.
Search Console · 16-month performance
38.2K clicks, 3.91M impressions, 1% CTR, average position 10. The spike in May–June 2025 is a one-off concert; traffic then settles to a lower, stable baseline. An honest read of this chart alone is “declining”.
Search Console · top queries
Hospital package pricing at 1,310 clicks, the concert at 1,002, a snake-catcher service at 410. These are the queries a resident actually types — none of them would appear on a content calendar.
Where 1,000 the top 1,000 exportable queries actually sit
Recomputed from the raw query export. Bands are disjoint and measured on each query's 16-month average position. Search Console caps both its list view and its CSV at 1,000 rows, so this is the distribution across the top 1,000 queries by clicks — a floor, not the site total.
113
Top 3 · 11.3% of the queries measured
833
Page one · 83.3% of the queries measured
Source: Google Search Console, coimbatorejunction.in, Search type: Web, 16 months ending 12 August 2026. Band thresholds are ours: #1 is an average below 1.5, since Search Console reports position as a decimal average rather than an integer rank.
Top queries by clicks · 16-month averages
| Query | Clicks | Impressions | CTR | Avg position |
|---|---|---|---|---|
| shanthi gears hospital full body checkup cost | 1,310 | 13,115 | 9.99% | 3.89 |
| ilayaraja concert coimbatore | 1,002 | 10,614 | 9.44% | 3.9 |
| snake catcher coimbatore | 410 | 3,196 | 12.83% | 3.3 |
| ilayaraja concert coimbatore 2025 | 274 | 2,014 | 13.60% | 3.41 |
| deva concert coimbatore | 268 | 3,107 | 8.63% | 3.1 |
| 111 bus route coimbatore | 266 | 2,668 | 9.97% | 6.11 |
| coimbatore ward number list pdf | 234 | 9,864 | 2.37% | 1.05 |
| shanthi gears hospital | 208 | 21,530 | 0.97% | 7.96 |
| post office selva magal scheme calculator | 187 | 2,426 | 7.71% | 5.84 |
| coimbatore dsp name list | 175 | 1,985 | 8.82% | 5.18 |
Top pages by clicks · Google, 16 months
| Page | Clicks | Impressions | CTR | Avg position |
|---|---|---|---|---|
| /coimbatore-ward-list/ | 5,714 | 401,847 | 1.42% | 6.27 |
| /shanthi-social-service-hospital-master-health-check-up/ | 5,243 | 266,993 | 1.96% | 7.62 |
| /event/ilaiyaraaja-truly-live-in-concert-2025/ | 3,329 | 44,415 | 7.50% | 7.87 |
| /111-bus-timings/ | 2,405 | 135,986 | 1.77% | 5.24 |
| /city-police-officers-list/ | 2,028 | 37,189 | 5.45% | 11.13 |
| /snake-catchers/ | 1,314 | 20,069 | 6.55% | 8.12 |
| /coimbatore-bus-routes/ | 1,045 | 232,588 | 0.45% | 11.15 |
| /tahsildar-office-coimbatore/ | 850 | 135,410 | 0.63% | 8.19 |
| /coimbatore-metro/ | 834 | 51,457 | 1.62% | 8.03 |
| /explore/red-taxi/ | 362 | 396,968 | 0.09% | 9.57 |
The last row is the red taxi guide: second-highest impressions on the site, 30th-ish by clicks. Judged on this table alone it is the weakest page here.
Search Console · pages, “1–10 of 1000”
That row counter is the reason this study never states a total page or query count. Search Console stops at 1,000 rows, so 1,000 is the export ceiling, not the number of pages ranking. Every share on this page is a share of that sample.
Evidence · AI search
The channel the clicks report does not contain
Both platforms that publish AI visibility data show the same direction on this property, over the same period the organic clicks were falling.
Google · 60,500 AI feature appearances, and climbing
Daily volume goes from roughly 400 in mid-May 2026 to about 1,700 by early August — a 4x climb in under three months. Top page: /explore/red-taxi/ at 36,386, then /coimbatore-ward-list/ at 11,529.
Bing · 40,101 Copilot citations, 35 pages a day
Citation sources: Microsoft Copilots and Partners. Over 180 days: 40,101 citations, averaging 223 a day, with 35 distinct pages cited on a typical day and a peak of 468 citations on 11 July 2026.
Bing Copilot citations per month
Aggregated from the 180-row daily export. Only complete calendar months are charted — plotting a 10-day August next to a 31-day July would draw a collapse that did not happen.
+45%
from Mar 2026 to Jul 2026, complete months only
34,280
citations across the 5 complete months charted
Excluded as partial periods: Feb 2026 (17 days, 3,848 citations); Aug 2026 (10 days, 1,973 citations). Source: Bing Webmaster Tools AI Performance daily export, 12 February – 10 August 2026. Bing notes that this report is a sample of overall activity, so the true totals are higher.
Two independent platforms, two different metrics, same direction. Google counts appearances inside its generative features; Bing counts times a page is quoted by Copilot. Neither is a click, and neither appears in a rankings report — which is precisely why this property looked like it was fading when it was actually being read more than ever.
Evidence · Bing
667.8K impressions on the engine behind Copilot
Bing is a rounding error in most traffic reports and the retrieval layer for Copilot, so on this property it is worth more than its click volume suggests.
Bing · 13.3K clicks, 667.8K impressions
Over 24 months at a 2% CTR. Note “events near me” at 228.1K impressions and 106 clicks — the same zero-click pattern as the Google side, on a generic query Bing answers in place.
Bing · pages, positions 4.3 to 5.5
The four top pages all average a better position on Bing than the site's Google average of 10. /coimbatore-ward-list/ converts at 10.41% here — the lookup page again outperforming on clicks because the answer needs the page.
Bing top keywords · 24 months
| Keyword | Impr | Clicks | Pos |
|---|---|---|---|
| events near me | 228.1K | 106 | 4.36 |
| coimbatore pincode | 18.8K | 43 | 3.95 |
| bookmyshow coimbatore | 8.8K | 10 | 6.24 |
| gold rate today in coimbatore | 4.6K | 48 | 5.27 |
Bing top pages · 24 months
| Page | Impr | CTR | Pos |
|---|---|---|---|
| /events/ | 232.0K | 0.15% | 4.31 |
| /coimbatore-pincode-list/ | 40.4K | 0.86% | 4.28 |
| /coimbatore-ward-list/ | 27.3K | 10.41% | 4.29 |
| /explore/broadway-cinemas/ | 24.8K | 0.97% | 5.46 |
Who it reaches
Four in five on a phone, 98% inside India
By device · Google
| Device | Clicks | Share | Avg pos |
|---|---|---|---|
| Mobile | 30,456 | 79.6% | 8.12 |
| Desktop | 7,664 | 20.0% | 18.13 |
| Tablet | 129 | 0.3% | 9.32 |
The mirror image of the Tamil property, which runs 36% desktop. Here it is a phone in the street looking up a bus, which is why the answer has to be at the top of the page.
By country · Google
| Country | Clicks | CTR | Avg pos |
|---|---|---|---|
| India | 37,478 | 1.04% | 9.05 |
| United Arab Emirates | 144 | 1.53% | 9.74 |
| United States | 119 | 0.08% | 16.01 |
| Singapore | 102 | 1.55% | 14.56 |
| Saudi Arabia | 43 | 1.10% | 13.94 |
| Qatar | 40 | 1.73% | 9.9 |
India is 98% of clicks. The Gulf and Singapore entries are the Coimbatore diaspora checking on things at home. The US line — 157,045 impressions, 119 clicks — is noise, and a traffic-first report would have called it international reach.
Coverage, stated honestly
What is proven here, and what is not
| Surface | Status | Finding | Evidence |
|---|---|---|---|
| Google organic | Documented | 38,249 clicks from 3.9M impressions over 16 months, at an average position of 10. 833 of the top 1,000 exportable queries sit on page one. | Search Console exports + screenshot |
| Google AI features | Documented | 60,500 appearances inside AI Overviews and AI Mode. Daily volume roughly quadrupled across the reporting window. | Search Console Generative AI report |
| Bing organic | Documented | 13,302 clicks from 665,490 impressions over 24 months, with four top pages averaging positions 4.3 to 5.5. | Bing Webmaster Tools export |
| Bing Copilot | Documented | 40,101 citations across 180 days, averaging 223 a day and 35 distinct pages cited. Monthly citations rose 45% from March to July 2026. | Bing AI Performance daily export |
| ChatGPT, Gemini, Perplexity | Not audited | No per-answer citation export exists for these surfaces, so this study does not measure or claim them. | No export available |
Why a city guide is in a SaaS SEO portfolio
Your docs are this site's red taxi page
Every SaaS site has pages with this exact profile: high impressions, low CTR, dense with single facts. The documentation. The integrations list. The pricing explainer. The glossary. The API reference.
On a standard rankings report those pages look like waste, and the standard recommendation is to consolidate, redirect or stop maintaining them. This property is where I learned to check the AI features report before making that call — because on a fact-dense page, a collapsing CTR is as likely to mean the answer is being served for you as it is to mean the page is failing.
The practical change to how I audit: every technical and content audit now pulls the Search Console generative-AI report and the Bing AI Performance report alongside the rankings data, and pages get classified as lookup (needs the click) or reference (wins the citation) before anyone judges their CTR. Applying one benchmark to both is how good pages get deleted.
Who did this work
Local writers, one operator
Alston Antony
Owner & SEO lead
CoimbatoreJunction.in is owned and operated by Alston Antony, who lives in the city and runs the SEO, the architecture and the maintenance schedule. The guides are written by residents rather than remote freelancers, because the details that make a local page trustworthy — which counter, what time it really opens, whether the route still runs — cannot be researched from somewhere else.
He has worked in search since 2010, holds an MSc in Computer Software Engineering (Distinction) from the University of Greenwich, and is a professional member of BCS, The Chartered Institute for IT. His day role is Senior Digital Marketing Manager at Brainstorm Force, whose software powers 7M+ sites.
Because the property is owned outright, there is no client approval step between the data and this page — which is why the declining organic chart is on it.
How to verify every number on this page
- Google totals come from Search Console for coimbatorejunction.in, Search type: Web, the 16-month window ending 12 August 2026. Site totals use the Devices export, which is complete; the query export is anonymised by Google and sums lower.
- The 1,000-row cap is real and stated. Search Console truncates both its list view and its CSV at 1,000 rows — visible as “1–10 of 1000” in the pages screenshot. Position bands are therefore shares of that sample, never of the site.
- Position bands were recomputed here from the raw export. “#1” means an average below 1.5, because Search Console reports position as a decimal average.
- Monthly AI citations were aggregated from the 180-row Bing daily export. February (17 days) and August (10 days) are partial and excluded from the chart, with their values printed underneath so nothing is hidden.
- Bing's own caveat is carried over: its AI Performance report states the data is a sample of overall activity, so 40,101 is a floor.
- Google and Bing AI figures are never added together. One counts appearances inside generative features, the other counts citations in Copilot answers. They are different metrics over different windows.
- The site is live. Open the red taxi guide and check the structure for yourself, then search a few of its pricing questions.
Case study FAQ