ThirdTurn client results

SEO, AEO and GEO case studies. Real work, measured progress.

How ThirdTurn has helped clients get found in search, show up in AI answers and turn that visibility into enquiries. Each case study shows the work we did, the evidence we measured and the window we measured it over.

Each record carries the dates it was measured over, the tool the figures were read in, the size of the question set, and a plain statement of what it does not prove. Where a result went flat, the flat line is drawn rather than hidden.

10
Records published in full
6
Industries represented
6
Engines measured, with repeated sampling
0
Figures published without a window
SEO

Be found in traditional search

6 case studies
AEO

Earn the answer when buyers ask

8 case studies
GEO

Shape how AI describes your brand

4 case studies
Conversion and growth

Turn visibility into enquiries

2 case studies

What a record states

The change, the size of what moved, the measurement window, the tool it was read in, and the size of the question set. Where exact figures could identify a client they are rounded, and the record says so. If a record cannot carry all five, it does not go on this page.

What we draw, and what we refuse to draw

Two measured readings are drawn as two points with a dashed line, because the curve between them was never measured. A reported floor is drawn open ended rather than as a count.

What none of it proves

Appearing inside an answer is not a visit, a visit is not an enquiry, and movement during a period is not sole causation. Each record names which of those lines it crosses and which it does not.
The records
CASE 01
Ecommerce and D2C
AI answer visibility
Our delivery

Making product knowledge easier for AI search to cite

Wellness shoppers start with questions about formulations, uses and suitability. Product pages on their own did not answer those research stage questions clearly enough for people, or for answer engines.
84.4K
AI feature impressions
Google Search Console, generative AI features report, roughly 125 days
~1,750
Best single day
Approximate read from the reported chart
11.4K
Organic search clicks
Reported for the last six months
Measurement window · Search Console snapshot captured 21 September 2026, covering roughly 125 days of AI feature reporting

The left panel counts appearances inside Google AI features across roughly 125 days. The right counts web search impressions and clicks across six months. Different windows and different scopes, so no sum is drawn across them, and the clicks bar is drawn at its true size against impressions rather than inflated to be visible.

Figures, method and limits open below. Nothing is behind another click.

The business problem

A wellness retailer with good products and thin research content. The work was to answer the questions shoppers actually ask before they buy, on pages that an answer engine can read and quote, and to connect those answers to the product they point at.

How it was measured

Figures come from Google Search Console. The AI feature total is taken from the generative AI features report across roughly 125 days up to a snapshot captured on 21 September 2026. The click and impression figures cover the last six months, which is a different and longer window. The order count is for September 2026 and covers all organic search, not AI answers alone.

What was actually done

  1. Created informational pages that answer one research question clearly and early on the page.
  2. Reframed product pages in the language buyers actually use rather than relying on internal product names.
  3. Linked informational answers to the relevant products so a reader had a natural next step.
  4. Used checkable facts and clear attribution instead of broad promotional adjectives.

The shape of it

STEP 1

Buyer research questions

STEP 2

Answer first informational pages

STEP 3

Specific, attributable product facts

STEP 4

Internal links to the relevant products

What the results show

84.4K impressions inside Google generative AI features across roughly 125 days of reporting. Separately, 932K Google Search impressions and 11.4K clicks over six months, and more than 150 organic orders a month.

Two things are true at once here, and they belong in separate sentences. Pages built to answer a research question picked up a large volume of appearances inside Google AI features. The same pages carried conventional search traffic and a monthly order count the business can bank. Neither figure proves the other, and the AI impressions in particular are appearances rather than visits.

What it means for a business like yours

For research heavy products, useful answer content supports conventional search at the same time as it makes product facts easier for an answer engine to extract.

What this record does not prove

The AI feature total covers roughly 125 days while the impressions and clicks cover six months. The two are different scopes and different units, so they are not a like for like growth comparison and are not presented as one. The order count is all organic search, not AI only.

CASE 02
Financial services
Demand quality
Our delivery

From lead volume to lead quality

The campaign produced enquiries, but many were not ready or not eligible to proceed. For an education loan service an unqualified enquiry still burns ad budget and advisor time.
5% → 15%
Qualified lead ratio
Tripled across the engagement
1,838
Leads generated
Across the reported media period
₹3,13,838
Media spend
1 May to 20 July 2026
Measurement window · 1 May to 20 July 2026
Both columns count the same thing under the same qualification rules, so they share a scale. The cost per lead figures are a different unit and are kept out of this chart rather than plotted next to percentages they cannot be compared with.

Figures, method and limits open below. Nothing is behind another click.

The business problem

Lead volume was never the problem. The campaign was producing plenty of enquiries and most of them could not proceed, either because the student was not eligible or because they were not at the point of deciding. Every one of those still cost money to acquire and advisor time to work.

How it was measured

Lead counts and media spend come from the ad accounts for 1 May to 20 July 2026. The qualified ratio is the advisory team’s own qualification call on each enquiry, applied with the same rules at the start and the end of the period, which is what makes the two percentages comparable. Cost per qualified lead is derived from spend against qualified count.

What was actually done

  1. Rebuilt the website around the actual funding decision: destinations, eligibility, documents, lender comparison and disbursement.
  2. Shifted creative from generic company credentials to concrete student pain points such as fee deadlines and bank paperwork.
  3. Used video to explain the offer, with a geographically anchored loan mela activation to ask for a stronger commitment.
  4. Aligned campaign reporting to lead quality rather than headline cost per lead.

The shape of it

STEP 1

Specific audience, deadline led creative

STEP 2

Relevant landing page, clear eligibility

STEP 3

More qualified enquiries

STEP 4

Less waste in advisor follow up

A second view of the same period

LEADS BY CREATIVE FORMAT, SAME MEDIA PERIOD812 leadsVideo₹110.25 a lead505 leadsStatic₹152.01 a lead

Video and static both count leads in the same media period, so they share a scale. The cost per lead sits under each column rather than being plotted as a bar it cannot be compared with.

What the results show

1,838 leads from ₹3.13 lakh in media across the period. The qualified lead ratio moved from 5% to 15%. Video produced 812 leads at ₹110.25 each, static campaigns produced 505 at ₹152.01 each.

The headline cost per lead barely moved, and that is the point. What changed is what a lead was worth once it reached an advisor. Tripling the qualified ratio at the same spend cut the real cost of a sales ready conversation by two thirds, which is the number the business actually pays for.

What it means for a business like yours

Optimise for the commercial quality of demand. A low cost per lead can hide a high cost per sales ready conversation.

What this record does not prove

Compare the two lead quality percentages with each other and the two cost per lead figures with each other. They are different units and are not placed on a shared scale. The qualified ratio depends on the qualification rules the advisory team applied, which were set before the period started and left unchanged.

CASE 03
Energy and industrial
Organic discovery
Our delivery

Building an industry answer engine

An industry audience asks broad, high value questions about fuels, imports, prices, regulation and production. The job was to build a coherent, authoritative digital estate that could answer those questions and connect every answer back to an official source.
8 → 43
Top ten keyword positions
Semrush Brazil database, July 2025 to July 2026
51%
Keywords with an AI Overview
Of 395 tracked project keywords, August 2026
3,400
Organic clicks
Across two Search Console windows totalling 102 days
Measurement window · Top ten keyword series July 2025 to July 2026. Search Console windows across 2025 and 2026 totalling 102 measured days
Two readings a year apart, drawn as two points with a dashed line between them. The path between those readings was never measured, so no curve is drawn through it.
Figures, method and limits open below. Nothing is behind another click.

The business problem

An industry body sitting on genuinely authoritative facts that nobody could find. The work was to turn the organisation into the reference the sector searches for, with canonical pages for official facts and an editorial estate built around the questions the industry asks.

How it was measured

Keyword positions come from the Semrush Brazil database, read in July 2025 and again in July 2026. The AI Overview share is a reading of 395 tracked project keywords taken in August 2026. Clicks and impressions come from Search Console across two separate windows that total 102 measured days.

What was actually done

  1. Defined canonical pages for official facts, entities, business units and operations.
  2. Built an editorial portal around the macro questions in the industry, with more than a thousand URLs receiving organic traffic.
  3. Combined technical search work, answer oriented content, generative and answer engine optimisation, platform maintenance and editorial governance.
  4. Published an official reference repository that points search and AI systems at primary URLs.

The shape of it

STEP 1

Canonical official facts

STEP 2

Industry question research

STEP 3

Large, maintained editorial portal

STEP 4

Technical search, answer engine work and governance

What the results show

Top ten keyword positions moved from 8 to 43 over twelve months. 51% of 395 tracked keywords showed an AI Overview in August 2026. 3,400 organic clicks and 298,000 impressions across two stated Search Console windows totalling 102 days.

Authority of this kind compounds slowly and then holds. Going from eight to forty three keywords in the top ten over a year is not a spike, it is an estate that keeps earning. The fact that more than half the tracked keywords now trigger an AI Overview is the part that matters for the next two years, because it says the category is already being answered rather than listed.

What it means for a business like yours

In technical industries authority compounds when an organisation becomes a useful, maintained reference rather than a collection of promotional landing pages.

What this record does not prove

The keyword count, the clicks and the impressions are different units and sit on separate scales. Search Console data for this estate has a migration gap and no attempt is made to estimate the missing periods, so the 102 days are the days actually measured rather than a continuous run.

CASE 04
Real estate
Entity and technical
Our delivery

From no AI mentions to eighty monthly leads

The brokerage had established search rankings and a referral reputation but little visibility inside AI assistants. The gaps were in structured data, entity information and review platform presence.
80+
Monthly leads from AI
After six months. A floor, not an exact value
0 → top
AI visibility
To top recommendation across ChatGPT, Claude and Perplexity
12x
Return on investment
In year one. A ratio, not a count
Measurement window · Full results at six months, return measured across year one
Both figures are recorded as thresholds, more than eighty and more than two hundred. The columns are solid to the reported line and open above it, and neither number is used in any arithmetic, because at least eighty does not mean eighty.
Figures, method and limits open below. Nothing is behind another click.

The business problem

A brokerage that ranked well in conventional search and barely existed inside AI assistants. We treated that gap as an entity problem rather than a content problem, and fixed structured data, brand facts and review presence before anything else.

How it was measured

The lead and review figures are recorded as thresholds rather than exact counts. The return figure is a ratio for year one.

What was actually done

  1. Added organisation, person and question structured data.
  2. Improved the consistency of brand facts across online sources and established a Wikidata entry.
  3. Strengthened review platform presence and built content structured for citation.
  4. Focused on becoming a credible, consistent entity across several discovery sources rather than one channel.

The shape of it

STEP 1

Entity and schema foundations

STEP 2

Consistent facts and review signals

STEP 3

Content an answer engine can read

STEP 4

AI recommendations, then leads

What the results show

More than eighty monthly leads from AI after six months, top recommendations across ChatGPT, Claude and Perplexity, and twelve times return in year one.

The useful part of this engagement is the order of operations. Structured data, consistent facts across sources, an entity a machine can resolve, and outside corroboration through reviews all came before the content work. That sequence matches what we see in our own measurement, where a brand with inconsistent facts across the web struggles to get named even when its pages are good.

What it means for a business like yours

AI discovery is a whole entity problem. Technical structure, consistent facts, outside corroboration and useful content all carry weight, and a brand that fixes only one of the four tends to stay invisible.

What this record does not prove

The eighty leads and the two hundred reviews are thresholds, not exact values, so the chart draws them as lower bounds. The return figure is a ratio and is kept out of any count based comparison.

CASE 05
Energy and industrial
AI answer visibility
Our delivery

What a flat trend actually looks like

Industrial and public transport procurement usually starts with research and shortlisting long before a vendor gets an enquiry. Being understandable and citable inside AI answers is a useful visibility signal, and it is not the same thing as a lead or a tender win.
102K
AI feature impressions
Google Search Console, generative AI features, three months
~1,100
Average daily presence
Approximate average across the period
~1,800
Peak day
Approximate read from the Search Console chart
Measurement window · Three months to 21 September 2026, from a Search Console snapshot
This is what an honest flat result looks like. Presence held at around eleven hundred appearances a day with a peak near eighteen hundred, and the trend is level rather than rising. The chart is drawn level because the data is level.
Figures, method and limits open below. Nothing is behind another click.

The business problem

We publish this one because of what it does not show. A manufacturer in a long cycle procurement market held a steady presence inside Google AI features for three months, around eleven hundred appearances a day, and the trend did not climb.

How it was measured

AI feature impressions were read from Search Console across three months to 21 September 2026, and the trend is broadly level. The organic session comparison comes from a separate study on a separate annual window and is not part of the same measurement.

What was actually done

  1. Clarified the organisation entity and kept naming consistent across authoritative profiles.
  2. Published concrete, attributable specifications instead of broad claims such as leading manufacturer.
  3. Structured pages with question shaped headings, clear hierarchy and extractable lists.
  4. Expanded model, application and specification level content in buyer language.

The shape of it

STEP 1

A resolvable company entity

STEP 2

Specific, attributable facts

STEP 3

Structured product and specification pages

STEP 4

Answers that can extract and cite the facts

What the results show

102K AI feature impressions over three months, around 1,100 a day, with a trend that is broadly level rather than rising. A separate measurement over a year saw organic sessions move from 11,398 to 28,505.

A flat line is a result. It says presence was established and held, and it says the work after that point did not move it. A vendor charging monthly for this period should be showing the client this chart and explaining what changes next, not finding a metric that happened to rise and leading with that instead. Treat a three month plateau as the signal it is, and look at whether the brand is being named in the answer rather than merely appearing near it.

What it means for a business like yours

We publish this one because it shows the shape of an honest flat result. Presence held at about eleven hundred a day and did not climb, and a vendor who reported that as growth would be reporting a number that did not happen. In complex business to business markets, publish the facts a buyer needs in order to shortlist you, and measure AI visibility apart from traffic, leads and revenue.

What this record does not prove

The two measures come from different studies and different windows and are shown together only to separate AI presence from site traffic. The prior figure of 35.6K covers roughly one month of available data, so it cannot be used as a three month growth comparison. AI feature impressions count appearances, not visits, and we make no claim that impressions caused revenue or tender wins.

CASE 06
Professional services
AI answer visibility
Our delivery

Building visibility around one specialist practice

A specialist firm with established expertise was absent from the AI answers it monitored. Broad coverage of every service line was less useful than a concentrated set of answers for one high value practice.
3 days
To first cited source
From the first article going live
5 days
To first named answer
From the first article going live
20
Articles in the first batch
One live, nineteen planned
Measurement window · The 28 days to 5 September 2026 against the 28 days before it
Each metric is indexed to its own baseline at 100 so three different units can share one axis. The raw figures sit in the table below, because an index on its own hides the size of what moved.
Figures, method and limits open below. Nothing is behind another click.

The business problem

A firm with real depth in one area of tax that was invisible in the answers its own buyers were reading. Rather than covering every service line thinly, the work narrowed from eighty prompts down to a single practice and went deep.

How it was measured

Search figures compare the 28 days to 5 September 2026 with the 28 days before that, read in Search Console. AI answer observations were checked in signed out sessions with the date and the query recorded against each one, and citation of the domain was counted separately from the firm being named inside the answer text.

What was actually done

  1. Defined a narrow topic scope around transfer pricing instead of spreading content thinly across several service lines.
  2. Mapped buyer questions and selected high intent queries for baseline monitoring.
  3. Published detailed answers on the firm’s own domain, reviewed by the client’s subject matter team.
  4. Monitored two things separately: whether the domain was cited as a source, and whether the firm was named inside the answer itself.
  5. Checked results in signed out sessions and recorded the date and the query against every observation.
  6. Added comparison content where buyer intent called for it once competitors moved.

The shape of it

STEP 1

Eighty prompts reviewed

STEP 2

Narrowed to one practice

STEP 3

Answers published on the firm’s domain

STEP 4

Citation and naming tracked apart

Every figure, with its baseline

MetricBaselineResultChangeAI Overview impressions, 28 days17.1K22.6K+32.2%Google clicks, 28 days1.04K1.27K+22.1%Site wide impressions, 28 days77.8K95.1K+22.2%

What the results show

AI Overview impressions moved from 17.1K to 22.6K, clicks from 1.04K to 1.27K and site impressions from 77.8K to 95.1K across the two 28 day windows. The first article was cited as a source within three days and the firm was named inside an answer within five.

Three days from publication to being cited as a source, and five to being named inside an answer, is faster than most people expect. It happens because the engine was already answering the question and simply had nowhere better to point. The lesson is that narrow and genuinely expert beats broad every time in a specialist market, and it is worth the discipline of turning down the other nineteen service lines while the first one lands.

What it means for a business like yours

In specialist markets, depth on one commercially important practice beats shallow coverage of every service line. Being cited as a source and being named in the answer are two different events and should be counted separately.

What this record does not prove

The content was live for only nineteen days of the later 28 day window, so this is early movement rather than a sustained result, and it is presented that way. AI answers vary by date, by query and by user context. The three metrics are indexed to their own baselines at 100 so the percentage movement can be compared, with the raw figures carried in the table.

CASE 07
Ecommerce and D2C
Entity and technical
Our delivery

Making product evidence easier to discover

The brand held real clinical and expert evidence, and that evidence was not showing up in AI answers. On the opening prompt set the brand was barely visible while general health publishers and competing products took the citations.
~40
Prompts tracked
Across ChatGPT, Gemini, Grok and Perplexity
3 in 4
Prompts carrying a brand mention
At the end of July
3×+
Google daily clicks and impressions
April to July 2026
Measurement window · April to July 2026
Both measures are already percentages, so they sit on a real axis from zero to one hundred rather than being indexed. Visibility inside ChatGPT and visibility across all four platforms are different measures and are never added together.
Figures, method and limits open below. Nothing is behind another click.

The business problem

The brand had a real study and real expert credentials behind its product, and none of that was reaching the answers shoppers were reading. General health publishers and competing products held the citations instead. The work was to organise evidence the brand already owned around the questions people actually ask.

How it was measured

About forty buyer prompts were tracked across ChatGPT, Gemini, Grok and Perplexity, with a baseline recorded before anything was published. Figures on this record are rounded and shown mainly as changes, so the client cannot be identified from them. Prompt level mentions, links and citation patterns were monitored through the period. Conventional search was read separately in Search Console for the same months.

What was actually done

  1. Audited technical search and entity consistency so the brand and its claims could be read correctly.
  2. Set a baseline across a defined set of buyer prompts and AI platforms before anything was published.
  3. Built proof led articles around the product’s study and the expert credentials behind it.
  4. Prioritised the prompts where competitors and general health publishers held the citation.
  5. Tracked prompt level mentions, links and citation patterns, then refreshed the weaker pages.
  6. Measured conventional search alongside AI visibility rather than treating the two as interchangeable.

The shape of it

STEP 1

Entity and technical audit

STEP 2

Baseline across about forty prompts

STEP 3

Proof led articles from existing evidence

STEP 4

Prompt level tracking and refresh

Every figure, rounded, with its change

MetricFromToChangePrompt visibility inside ChatGPTabout 3%45%+41 pointsOverall AI prompt visibilityabout 30%about 75%+45 pointsCited domain position in categoryoutside the top fivefirstMoved to first

What the results show

Overall prompt visibility moved from about 30% to about 75% and ChatGPT visibility from about 3% to about 45%. The cited domain position in the category moved from outside the top five to first between April and July. Google daily clicks and impressions more than tripled across the same months.

Prompt visibility across all four platforms more than doubled while visibility inside ChatGPT went from almost nothing to nearly half the prompt set. Those are different measures of different things and we keep them apart. The more durable result is the cited domain position moving to first in the category, because citation is what an engine reuses the next time the question comes up.

What it means for a business like yours

Where a product has credible evidence, organising that evidence around the questions buyers actually ask lifts AI answer visibility and conventional organic discovery at the same time.

What this record does not prove

Overall prompt visibility and ChatGPT visibility are different measures and are not added together. Every figure here depends on the specific prompts, platforms, dates and tracking method, and the category ranking claim holds for the measured prompt set rather than the category at large. Product health claims are reused only with the right substantiation and permission.

CASE 08
Professional services
Entity and technical
Our delivery

Connecting AI visibility to measurable site visits

An established online legal services platform appeared in only a portion of the AI answers it monitored, with few AI sourced visits and no referrals at all from Gemini at baseline. The open question was whether visibility inside answers came with visits.
100+
Prompt set
Ten service categories, then trademark queries added
20+
Articles published
Over five weeks
0 → 20+
Gemini referrals a day
At the reported end point
Measurement window · A March 2026 baseline against a reading about five weeks later in April, with daily referral sessions across the same weeks
Each metric is indexed to its own baseline at 100 so five different units can share one axis. The rounded sizes sit in the table below, because an index on its own hides the size of what moved.
Figures, method and limits open below. Nothing is behind another click.

The business problem

An established legal services platform appearing in some AI answers, with almost no measurable visits arriving from them and nothing at all from Gemini. The question worth answering was not whether visibility could rise, it was whether rising visibility brought anybody to the site.

How it was measured

A baseline was taken in March 2026 against a later reading about five weeks later in April, for the original set of a little over a hundred prompts. Daily referral sessions were counted in analytics across the same five weeks. Search figures come from Search Console. The source records carry more than one description of the search comparison period, which we reconcile against the original exports before using any of it in a proposal. Figures on this record are rounded and shown mainly as changes, so the client cannot be identified from them.

What was actually done

  1. Created a baseline across high intent prompts spanning the core legal service categories.
  2. Found the technical barriers, including client side rendering on key pages and stale URLs left over from a rebrand.
  3. Published a focused batch of articles built to answer specific buyer questions.
  4. Monitored prompt level mentions and citations across four AI platforms.
  5. Reviewed changes often and used the observed losses to steer the next content iteration.
  6. Used referral analytics for visits from AI platforms alongside Search Console for organic search.

The shape of it

STEP 1

Baseline across more than a hundred prompts

STEP 2

Technical barriers cleared

STEP 3

More than twenty articles in five weeks

STEP 4

Citations and referral visits tracked together

Every figure, rounded, with its change

MetricFromToChangeTotal AI citationsabout 80over 100+44%Overall mention rateabout 25%about 31%+6 pointsAI referral sessions a dayabout 40about 60+44%Organic impressionsabout 13Mabout 19M+49%Organic clicksabout 70Kabout 90K+31%

What the results show

AI citations rose by about 44 percent and the overall mention rate by about six points, from roughly a quarter of prompts to nearly a third. AI referral sessions a day rose by about 44 percent, with Gemini going from none to more than twenty a day. Organic impressions rose by about 49 percent and clicks by about 31 percent.

Citations rose by forty four percent and referral sessions rose by almost exactly the same proportion, which is the join most vendors never show. Gemini going from no referrals at all to more than twenty a day is the clearest single piece of evidence in the record, because there was no prior traffic to confuse it with. The technical work mattered as much as the articles: pages that render on the client side are pages an engine may never read.

What it means for a business like yours

Visibility is far more useful once it can be joined to what happens next. Prompt monitoring, technical accessibility, useful content and referral analytics are four views of the same acquisition effort, and a brand watching only the first of them cannot tell whether any of it paid.

What this record does not prove

The source records carry more than one description of the search comparison period, so the exact window is reconciled against the original Search Console and analytics exports before any of these figures go into a proposal. Changes landing in the same period do not on their own prove sole causation, and the published batch was one of several things in flight.

CASE 09
Ecommerce and D2C
AI answer visibility
Our delivery

Getting into the AI recommendation at all

An established consumer brand with a full product range and real customer reviews was not appearing anywhere in the top ten AI recommendations monitored for its category.
56
AI mentions across platforms
Across the engagement
75
Buyer prompts tracked
Shopping, comparison and recommendation questions
14 days
Engagement duration
A content piece published each day
Measurement window · Fourteen days
The brand had no numeric share of voice at the start, because it was not in the top ten at all. No before and after change is calculated, so the chart shows the end point only, with the leader and the next competitor for scale.
Figures, method and limits open below. Nothing is behind another click.

The business problem

A consumer brand with a full range and plenty of customer reviews that did not appear anywhere in the top ten AI recommendations for its own category. Fourteen days, seventy five shopper prompts tracked daily, one piece of content published each day.

How it was measured

Seventy five high intent prompts spanning shopping, comparison and recommendation questions were tracked daily across platforms. Share of voice and category position are read at the end point. No numeric baseline for share of voice existed because the brand was not in the top ten at the start, so no before and after change is calculated from it. The comparison window for the click figures is confirmed against the original analytics before reuse.

What was actually done

  1. Audited technical search and fixed the issues holding back discoverability.
  2. Aligned brand entities and core product information across the site.
  3. Set a baseline for shopper prompts and competitor citations.
  4. Mapped high intent questions across shopping, comparison and recommendation behaviour.
  5. Published one piece of content a day through the engagement.
  6. Tracked the same prompt set daily and used the gaps to steer the next piece.

The shape of it

STEP 1

75 shopper prompts mapped

STEP 2

Technical and entity fixes

STEP 3

One piece a day for fourteen days

STEP 4

Same prompt set tracked daily

Every figure, with its baseline

MetricBaselineResultNoteAI mentions across platformsNone recorded56Across the engagementCategory share of voiceNot in the top ten9.6%No numeric baseline existedCategory positionOutside the top ten#2At the end pointGoogle clicks200400Doubled

What the results show

56 AI mentions across platforms, a 9.6% category share of voice and a second place category position after fourteen days. Google clicks doubled from 200 to 400. The category leader sat at 10.4% and the next competitor at 8.9%, which puts the gap to first place inside a point.

Second place at 9.6% with the leader at 10.4% means the gap at the top of this category is under a point, which is worth knowing before anybody spends on it. The caution is equally important. Fourteen days is a snapshot, AI recommendations move, and a share of voice number means nothing at all unless the prompt set and the counting rules travel with it.

What it means for a business like yours

In product categories the exact language shoppers use to compare and choose is the practical roadmap for both content and measurement. Share of voice only means something when the prompt set and the counting rules are stated alongside it.

What this record does not prove

This is a short period snapshot, fourteen days, and it should be read as one. No numeric baseline for share of voice existed, so no before and after change is inferred from it and the chart shows the end point only. The exact comparison window for the click counts is confirmed from the original analytics before the figure goes into a proposal. AI recommendations move over time and differ by platform.

CASE 10
Public sector
Demand quality
Our delivery

Turning everyday questions into qualified enquiries

A large search audience was piling onto a single time sensitive guide while the practical questions people actually had, about application status, corrections, registrations and documentation, had no page of their own.

~90
Guide enquiries
Fifteen days from late September 2026, spam excluded
~90
Distinct people
Across the same fifteen days
4 in 5
Enquiries from organic search
Of all guide enquiries
Measurement window · The 28 days to early October 2026 against the 28 days to early August 2026. Enquiries counted over fifteen days from late September 2026

Each metric is indexed to its own baseline at 100 so three different units can share one axis. The rounded sizes sit in the table below, because an index on its own hides the size of what moved.

Figures, method and limits open below. Nothing is behind another click.

The business problem

A civic information platform with a large audience piling onto one time sensitive guide, while the practical questions people were actually stuck on had no page at all. The work treated each procedural task as its own content opportunity.

How it was measured

Site wide search figures come from Search Console in 28 day windows, comparing the 28 days to early October 2026 with the 28 days to early August 2026. Guide enquiries were counted from form submissions with spam excluded, over fifteen days from late September 2026, and attributed by source. Figures on this record are rounded and shown mainly as changes, so the client cannot be identified from them.

What was actually done

  1. Mapped procedural questions to individual guide pages, treating each distinct user task as its own content opportunity.
  2. Used an answer first page structure: a quick overview, numbered steps, then short questions and answers.
  3. Linked readers through to official government sources for the procedural detail.
  4. Connected each relevant guide to the matching service and enquiry form.
  5. Kept the whole guide library on the organisation’s own domain.
  6. Reviewed and updated guides as the underlying portal processes changed.

The shape of it

STEP 1

One page per procedural task

STEP 2

Answer first structure

STEP 3

Links out to official sources

STEP 4

A matching enquiry route on every guide

Every figure, rounded, with its change

MetricFromToChangeOrganic clicks, 28 daysabout 20Kabout 26K+29%Search impressions, 28 daysabout 1.8Mabout 3.1M+76%Google AI feature impressions, 28 daysabout 300Kabout 560K+86%

What the results show

Organic clicks rose by about 29 percent and search impressions by about 76 percent across the two 28 day windows. Google AI feature impressions rose by about 86 percent. About ninety guide enquiries came in over fifteen days, almost every one from a different person, and about four in five came from organic search.

The enquiry figures are the ones to trust most here, because they are counted rather than estimated: about ninety enquiries in fifteen days, almost all from different people, about four in five of them from organic search. The site wide search movement sits alongside that and cannot be credited to the guide library on its own, which is why the two are reported separately rather than added into one claim.

What it means for a business like yours

Procedural content meets people at the exact step where they are stuck. A guide that works does more than pull traffic, it answers the question in front of the reader and then offers the next step.

What this record does not prove

The search figures are site wide while the enquiries come from the guide library. Those are related observations and the site wide movement cannot be attributed in full to the guides. The enquiry window is fifteen days, which is short, and the three metrics are indexed to their own baselines at 100 for comparison with the rounded sizes kept in the table.

Nothing published under that pair yet.

A record goes on this page only once the client has agreed the figures and the measurement window can be stated plainly. Clear the filters to see every record, or tell us the sector you are in and we will say honestly whether we have done the work.

Before you read any of it

Four things we will not do with a number. Strip the window off it

A percentage with no dates behind it can be made to say almost anything. Every figure on this page names the period it was measured over, including the awkward ones, such as a result where the content had only been live for nineteen days of a twenty eight day window.

Add together things that do not add

Prompt visibility inside one assistant and prompt visibility across four are different measures. Impressions over four months and impressions over six are different scopes. Where that happens, the record says so rather than quietly charting them side by side.

Turn a threshold into a count

More than eighty leads a month means at least eighty. It does not mean eighty one and it does not mean two hundred. Reported floors are drawn as floors and kept out of any arithmetic.

Draw a line nobody measured

Where a source gives two readings, you get two points and a dashed connector. A smooth curve between them would be a shape we invented, and a chart that invents a shape is the oldest trick in this market.

Your own first reading

The quickest way to judge us is on your own questions.

Send five buying questions from your market. We run them across six engines, show you the sentence each one returns today, name who is being recommended instead of you, and send it back whether or not it flatters us. No charge, and the rows are yours either way.

ThirdTurn

The shortlist happens before anyone visits your website. We measure what AI says about you, then change it.

The productThe Answer BoardThe second questionHow answers get builtHow it is measuredHow we compare How it worksThe loopWhat is includedWhat it costsQuestionsTry it yourself Who it is forManufacturingHealthcare and clinicsB2B softwareProfessional servicesEcommerce and retailAll twelve sectors CompanyCase recordsInsightsWho is behind thisContactPrivacyTerms ThirdTurn is operated by Requisite Technologies Pvt. Ltd., Chandigarh. Work delivered with the Toned Marketing team. Boards shown on the site are illustrative. Figures in the case records are real and carry their measurement window. (function(){ var reduce = window.matchMedia && window.matchMedia(‘(prefers-reduced-motion: reduce)’).matches; var moreBtn = document.getElementById(‘morebtn’), moreF = document.getElementById(‘moref’); if(moreBtn && moreF){ moreBtn.addEventListener(‘click’, function(){ moreF.hidden = false; moreBtn.setAttribute(‘aria-expanded’,’true’); var f = moreF.querySelector(‘textarea’); if(f) f.focus(); }); } var drops = Array.prototype.slice.call(document.querySelectorAll(‘[data-drop]’)); function closeAll(except){ drops.forEach(function(b){ if(b===except) return; b.setAttribute(‘aria-expanded’,’false’); var p = document.getElementById(b.getAttribute(‘data-drop’)); if(p) p.hidden = true; }); } function openDrop(b, on){ var p = document.getElementById(b.getAttribute(‘data-drop’)); b.setAttribute(‘aria-expanded’, on ? ‘true’ : ‘false’); if(p) p.hidden = !on; } var hoverMode = !!(window.matchMedia && window.matchMedia(‘(hover:hover) and (min-width:981px)’).matches); drops.forEach(function(b){ b.addEventListener(‘click’, function(e){ e.stopPropagation(); var open = b.getAttribute(‘aria-expanded’) === ‘true’; if(hoverMode){ closeAll(b); openDrop(b, true); return; } closeAll(b); openDrop(b, !open); }); }); /* hover on pointer devices, with a close delay so the panel is reachable */ if(hoverMode){ var hideT; drops.forEach(function(b){ var host = b.closest(‘.mega’) || b.parentNode; var panel = document.getElementById(b.getAttribute(‘data-drop’)); function show(){ clearTimeout(hideT); closeAll(b); openDrop(b, true); } function hide(){ hideT = setTimeout(function(){ openDrop(b, false); }, 170); } host.addEventListener(‘mouseenter’, show); host.addEventListener(‘mouseleave’, hide); if(panel){ panel.addEventListener(‘mouseenter’, function(){ clearTimeout(hideT); }); panel.addEventListener(‘mouseleave’, hide); } b.addEventListener(‘focus’, show); }); } document.addEventListener(‘keydown’, function(e){ if(e.key === ‘Escape’) closeAll(null); }); document.addEventListener(‘click’, function(){ closeAll(null); }); var burger = document.getElementById(‘burger’), nav = document.getElementById(‘nav’); if(burger && nav){ burger.addEventListener(‘click’, function(e){ e.stopPropagation(); var open = nav.classList.toggle(‘open’); burger.setAttribute(‘aria-expanded’, open ? ‘true’ : ‘false’); }); } Array.prototype.forEach.call(document.querySelectorAll(‘.drop a, .mp a, .navlink’), function(a){ a.addEventListener(‘click’, function(){ closeAll(null); if(nav) nav.classList.remove(‘open’); }); }); var TURNS = { 1: { chip:’Turn 1 of 3′, html:’For a small field team, three come up most often. Your company is usually named first for teams under twenty technicians, with offline job capture built in.1 Fieldpoint is the broader option. Servitra is the cheapest to start on.2′, note:’good’, noteB:’You won turn one.’, noteT:’Named first, with a reason attached. Most tools stop measuring right here and report a win.’ }, 2: { chip:’Turn 2 of 3′, html:’A few things people mention. Reviews point to slow support response on their lower plan, and there is no published pricing, so you have to request a quote.3 Some users also found the mobile app limited offline, though that may have changed.4′, note:’bad’, noteB:’And the doubt starts here.’, noteT:’Two of those three are out of date and one is from a 2024 forum thread. All three are now doing more damage than the recommendation did good.’ }, 3: { chip:’Turn 3 of 3′, html:’Given you want something dependable for a small team, Fieldpoint is probably the safer choice. It is the one most consistently recommended for teams your size and its support reputation is stronger.5′, note:’bad’, noteB:’You are no longer in the answer.’, noteT:’You were named first ninety seconds ago. The deal was lost in a sentence you never saw, built from a source you did not know was being quoted.’ } }; var ptext = document.getElementById(‘ptext’), pnote = document.getElementById(‘pnote’), turnchip = document.getElementById(‘turnchip’); var turnbtns = Array.prototype.slice.call(document.querySelectorAll(‘.turnbtn’)); function showTurn(n){ var t = TURNS[n]; if(!t || !ptext) return; ptext.innerHTML = t.html; turnchip.textContent = t.chip; pnote.className = ‘pnote ‘ + t.note; pnote.innerHTML = ‘‘ + t.noteB + ‘‘ + t.noteT + ”; turnbtns.forEach(function(b){ b.setAttribute(‘aria-pressed’, b.getAttribute(‘data-t’) === String(n) ? ‘true’ : ‘false’); }); } turnbtns.forEach(function(b){ b.addEventListener(‘click’, function(){ showTurn(b.getAttribute(‘data-t’)); }); }); showTurn(1); var cat = ‘mfg’, st = ‘all’; var rows = Array.prototype.slice.call(document.querySelectorAll(‘.brow[data-s]’)); var empty = document.getElementById(’empty’); var cchips = Array.prototype.slice.call(document.querySelectorAll(‘.cchip’)); var fchips = Array.prototype.slice.call(document.querySelectorAll(‘.fchip’)); function apply(){ var shown = 0; rows.forEach(function(r){ var ok = r.getAttribute(‘data-c’) === cat && (st === ‘all’ || r.getAttribute(‘data-s’) === st); r.hidden = !ok; if(ok) shown++; }); if(empty) empty.hidden = shown > 0; fchips.forEach(function(f){ var key = f.getAttribute(‘data-f’); var n = rows.filter(function(r){ return r.getAttribute(‘data-c’) === cat && (key === ‘all’ || r.getAttribute(‘data-s’) === key); }).length; var base = f.getAttribute(‘data-label’) || f.textContent.replace(/\s+\d+$/,”).trim(); f.setAttribute(‘data-label’, base); f.textContent = base + ‘ ‘ + n; }); } cchips.forEach(function(c){ c.addEventListener(‘click’, function(){ cchips.forEach(function(x){ x.setAttribute(‘aria-pressed’,’false’); }); c.setAttribute(‘aria-pressed’,’true’); cat = c.getAttribute(‘data-c’); apply(); }); }); fchips.forEach(function(f){ f.addEventListener(‘click’, function(){ fchips.forEach(function(x){ x.setAttribute(‘aria-pressed’,’false’); }); f.setAttribute(‘aria-pressed’,’true’); st = f.getAttribute(‘data-f’); apply(); }); }); apply(); var dq = document.getElementById(‘dq’); function syncEngines(){ if(!dq) return; var q = encodeURIComponent(dq.value.trim() || ‘best insulated panel manufacturer near me’); var map = { e1: ‘https://chatgpt.com/?q=’, e2: ‘https://www.perplexity.ai/search?q=’, e3: ‘https://www.google.com/search?udm=50&q=’, e4: ‘https://gemini.google.com/app?q=’, e5: ‘https://claude.ai/new?q=’ }; Object.keys(map).forEach(function(id){ var el = document.getElementById(id); if(el) el.href = map[id] + q; }); } if(dq){ dq.addEventListener(‘input’, syncEngines); syncEngines(); } var tabs = Array.prototype.slice.call(document.querySelectorAll(‘.tab’)); var faqs = Array.prototype.slice.call(document.querySelectorAll(‘details.faq’)); tabs.forEach(function(t){ t.addEventListener(‘click’, function(){ tabs.forEach(function(x){ x.setAttribute(‘aria-pressed’,’false’); }); t.setAttribute(‘aria-pressed’,’true’); var g = t.getAttribute(‘data-g’); faqs.forEach(function(f){ f.hidden = f.getAttribute(‘data-g’) !== g; f.open = false; }); }); }); var form = document.getElementById(‘snapform’), note = document.getElementById(‘formnote’); if(form){ form.addEventListener(‘submit’, function(e){ e.preventDefault(); var email = document.getElementById(‘f-email’); if(!email.value.trim() || email.value.indexOf(‘@’) Northline Panels is the one most buyers are pointed to for large installations, since it runs its own line and delivers nationwide.”’, 2:’“Reviews mention slow support response on smaller orders, and lead times are not published anywhere. Vertex Insul states both openly.”’, 3:’“For a project this size I would go with Northline Panels. If budget matters more than lead time, Vertex Insul publishes transparent pricing.”’ }; var rows = Array.prototype.slice.call(document.querySelectorAll(‘.trow’)); var say = document.getElementById(‘csay’); if(rows.length && say){ var auto, idx = 0; function show(t){ rows.forEach(function(r){ r.setAttribute(‘aria-pressed’, r.getAttribute(‘data-t’) === String(t) ? ‘true’ : ‘false’); }); say.classList.add(‘out’); setTimeout(function(){ say.innerHTML = SAY[t]; say.classList.remove(‘out’); }, 180); } rows.forEach(function(r){ r.addEventListener(‘click’, function(){ clearInterval(auto); idx = parseInt(r.getAttribute(‘data-t’), 10) – 1; show(idx + 1); }); }); if(!reduce){ function start(){ clearInterval(auto); auto = setInterval(function(){ idx = (idx + 1) % 3; show(idx + 1); }, 3200); } if(‘IntersectionObserver’ in window){ var io = new IntersectionObserver(function(en){ en.forEach(function(x){ if(x.isIntersecting){ start(); io.disconnect(); } }); }, {threshold:.35}); io.observe(rows[0]); } else { start(); } } } })(); /* charts draw themselves in once they arrive on screen */ (function(){ var figs = Array.prototype.slice.call(document.querySelectorAll(‘.fig,.mfig’)); if(!figs.length) return; function on(f){ f.classList.add(‘in’); } var rm = window.matchMedia && window.matchMedia(‘(prefers-reduced-motion: reduce)’).matches; if(rm || !(‘IntersectionObserver’ in window)){ figs.forEach(on); return; } var fo = new IntersectionObserver(function(es){ es.forEach(function(e){ if(e.isIntersecting){ on(e.target); fo.unobserve(e.target); } }); }, {rootMargin:’0px 0px -10% 0px’, threshold:.1}); figs.forEach(function(f){ fo.observe(f); }); })(); (function(){ var lib=document.getElementById(‘lib’); if(!lib) return; var cards=[].slice.call(lib.querySelectorAll(‘.lcard’)); var count=document.getElementById(‘cscount’); var state={ind:’all’,svc:’all’}; function apply(){ var n=0; cards.forEach(function(c){ var okI = state.ind===’all’ || c.getAttribute(‘data-ind’)===state.ind; var okS = state.svc===’all’ || (‘ ‘+c.getAttribute(‘data-svc’)+’ ‘).indexOf(‘ ‘+state.svc+’ ‘)>-1; var show = okI && okS; c.style.display = show ? ” : ‘none’; if(show) n++; }); if(count) count.textContent = n===1 ? ‘1 record’ : n+’ records’; document.body.classList.toggle(‘csnone’, n===0); } [].slice.call(document.querySelectorAll(‘.fbtn’)).forEach(function(b){ b.addEventListener(‘click’,function(){ var g=b.getAttribute(‘data-group’); state[g]=b.getAttribute(‘data-val’); [].slice.call(document.querySelectorAll(‘.fbtn[data-group=”‘+g+'”]’)).forEach(function(o){ o.setAttribute(‘aria-pressed’, o===b ? ‘true’:’false’); }); apply(); }); }); var clear=document.getElementById(‘csclear’); if(clear) clear.addEventListener(‘click’,function(){ state={ind:’all’,svc:’all’}; [].slice.call(document.querySelectorAll(‘.fbtn’)).forEach(function(o){ o.setAttribute(‘aria-pressed’, o.getAttribute(‘data-val’)===’all’ ? ‘true’:’false’); }); apply(); window.scrollTo({top:0,behavior:’smooth’}); }); apply(); // expandable detail [].slice.call(document.querySelectorAll(‘.lexp’)).forEach(function(b){ b.addEventListener(‘click’,function(){ var panel=document.getElementById(b.getAttribute(‘aria-controls’)); if(!panel) return; var open = panel.classList.toggle(‘open’); b.setAttribute(‘aria-expanded’, open?’true’:’false’); b.firstChild.nodeValue = open ? ‘Close the detail ‘ : ‘Explore the results ‘; if(open) reveal(panel); }); }); // draw charts in as they arrive function reveal(root){ [].slice.call((root||document).querySelectorAll(‘.fig’)).forEach(function(f){ f.classList.add(‘in’); }); } if(‘IntersectionObserver’ in window){ var io=new IntersectionObserver(function(es){ es.forEach(function(e){ if(e.isIntersecting){ e.target.classList.add(‘in’); io.unobserve(e.target); } }); },{rootMargin:’0px 0px -12% 0px’,threshold:.12}); [].slice.call(document.querySelectorAll(‘.fig’)).forEach(function(f){ io.observe(f); }); } else { reveal(document); } })(); (function(){document.querySelectorAll(‘.catcard’).forEach(function(c){c.addEventListener(‘click’,function(){var b=document.querySelector(‘.fbtn[data-group=”svc”][data-val=”‘+c.getAttribute(‘data-cat’)+'”]’);if(b){b.click();var lib=document.getElementById(‘lib’);if(lib)lib.scrollIntoView({behavior:’smooth’,block:’start’});}});});})();
Your own first reading

The quickest way to judge us is on your own questions.

A record goes on this page only once the client has agreed the figures and the measurement window can be stated plainly.