NEWHow a B2B SaaS went from 4 to 131 AI answers citing it
TechDevs
ContactBook a free call

Case study

131 of 214 prompts now cite the brand. It started at 4.

Ranked inside ChatGPT before their competitors knew it was a channel

Book your free strategy callRead the strategy
131/214target AI prompts now cite the brand (from 4)

Prompts citing the brand (of 214 tracked)

Client
B2B compliance SaaS · US
Engagement
AI SEO · Generative Engine Optimization
Timeframe
8 months tracked
Led by
Prithvi Jadwani

The outcome

Where it started. Where it is now.

MeasureBeforeAfter
Buying-intent prompts citing the brand4 of 214131/214
Demo requests from AI assistantsFlatlined19%
Conversion rate, indexed to classic organic1×4.4×for visitors from AI assistants

The situation

Rankings were fine. The buyers had left the page.

Organic rankings were stable, but demo requests had flatlined. Their buyers had moved: prospects were asking ChatGPT and Perplexity "best SOC 2 compliance tools" and the brand appeared in none of those answers.

What we found

  1. 01

    Prospects were asking ChatGPT and Perplexity for a shortlist of SOC 2 tools, not searching Google for one.

  2. 02

    Across 214 buying-intent prompts on three assistants, the brand was named in 4.

  3. 03

    The pages that ranked were written for a crawler that ranks, not a model that quotes. Nothing on them was easy to lift into an answer.

Citation audit · yourbrand.comBaseline

214 buying-intent prompts

ChatGPT, Perplexity, Gemini

Cited in0
Cited Not cited0 / 214 run

“best SOC 2 compliance tools”

Named: Competitor A, B, C

Not cited

“SOC 2 software for a small startup”

Named: Competitor B, A

Not cited

Cell positions are illustrative · example prompts are placeholders

The strategy

Optimise to be cited, not just to rank.

The obvious fix was more blog posts. That would have chased a Google position the brand already held. The real gap was upstream: assistants were building the shortlist before a buyer ever clicked. So the plan treated every buying question as a slot in an AI answer, and worked on the three things that decide who fills it: what the page says, who else vouches for it, and whether the model can read it at all.

  1. Starting point

    Rankings were fine.

  2. Move 01 · Measure

    Score every buying question on every assistant

    • 214 buying-intent prompts, three assistants
    • Competitors named in each answer, logged
    • Baseline: cited in 4

  3. Move 02 · Rewrite

    Rebuild key pages around answers a model can lift

    • Definition in the first lines
    • Comparison tables in place of prose
    • Stats carry a named source

  4. Move 03 · Distribute

    Get named where the models already look

    • Review sites
    • Reddit threads
    • Industry roundups

  5. Move 04 · Unblock and track

    Let AI crawlers in, then watch the share weekly

    • llms.txt shipped
    • GPTBot and PerplexityBot allowed
    • Citation share tracked weekly

  6. Outcome

    131/214

    target AI prompts now cite the brand (from 4)

Execution

What we did, move by move.

Each move below is a piece of the strategy above, with a drawing of the work as it ran.

Move 01 · Measure

Score every buying question on every assistant

We wrote down the 214 prompts a compliance buyer actually asks, ran each one through ChatGPT, Perplexity and Gemini, and logged who got named. That table became the scoreboard for the whole engagement.

  • 214 buying-intent prompts, three assistants
  • Competitors named in each answer, logged
  • Baseline: cited in 4
Prompt audit sheet · yourbrand.com214 prompts · 3 assistants
Buying promptChatGPTPerplexityGemini
“best SOC 2 compliance tools”
“SOC 2 software for a 50-person startup”
“alternatives to Competitor A”
“fastest way to pass a SOC 2 audit”
ABC Competitors named in the answeryourbrand not named

Illustration of the method · prompts and names are placeholders

Move 02 · Rewrite

Rebuild key pages around answers a model can lift

Assistants quote the clearest sentence on the page. We moved definitions to the top, turned feature prose into comparison tables, and backed every claim with a stat and a named source.

  • Definition in the first lines
  • Comparison tables in place of prose
  • Stats carry a named source

BeforeFeature prose

Nothing short enough to quote

AfterBuilt to be lifted

What is SOC 2 compliance software?

DefinitionSOC 2 compliance software collects audit evidence and monitors controls so a team can pass its SOC 2 audit.

Comparison table
yourbrandComp. A
Evidence collection
Control monitoring
Auditor access
Stat + named source

[stat] claim, stated once, with where it comes from.

Source: [named report]

Illustration of the page pattern · copy is a placeholder

Move 03 · Distribute

Get named where the models already look

A model trusts a brand that other sources mention. We placed the brand in the review sites, Reddit threads and industry roundups that assistants cite for this category, and added structured data so the facts matched everywhere.

  • Review sites
  • Reddit threads
  • Industry roundups
  • Structured data

Illustration of the method · sources and answer are placeholders

Move 04 · Unblock and track

Let AI crawlers in, then watch the share weekly

None of it counts if the bots cannot read the site. We shipped llms.txt, fixed crawler access for GPTBot and PerplexityBot, and re-ran the prompt set every week so each change could be tied to a move in citation share.

  • llms.txt shipped
  • GPTBot and PerplexityBot allowed
  • Citation share tracked weekly
robots.txtllms.txtyourbrand.com
 # yourbrand.com/robots.txt User-agent: GPTBot Disallow: /   User-agent: PerplexityBot Disallow: /

Citation share re-runEvery Monday · 214 prompts · 3 assistants

Weekly
  1. Wk 1
  2. Wk 2
  3. Wk 3
  4. Wk 4
  5. Wk 5
  6. Wk 6
  7. Wk 7
  8. Wk 8

Illustration · file contents are placeholders

How it ran

The engagement, in order.

  1. 1

    Phase 1

    Baseline

    Prompt set written, run across three assistants, and scored. Cited in 4 of 214.

  2. 2

    Phase 2

    Fix the pages

    Key pages rebuilt around definitions, tables and sourced stats. Crawler access fixed.

  3. 3

    Phase 3

    Earn the mentions

    Brand placed in the review sites, threads and roundups the assistants cite.

  4. 4

    Phase 4

    Compound

    Weekly re-runs steer the next page and the next placement. Eight months tracked.

The evidence

Citations climbed every month for eight months.

Each point is the same 214 prompts, re-run. No month went backwards.

Prompts citing the brand (of 214 tracked)

Same prompt set, re-run every month

050100150214214 prompts trackedM1M2M3M4M5M6M7M8921386289104Start · 4131of 214M8 · prompts citing the brand050100150214214 prompts trackedM1M2M3M4M5M6M7M8Start · 4131of 214M8 · cited
M1

4of 214 cite the brand

Cites the brand Does not yet
target AI prompts now cite the brand (from 4)
131/214
of demo requests now arrive from AI assistants
19%
AI visitors convert vs classic organic visitors
4.4×

What carries over

Three things to take into your own business.

  1. 01

    Rank and citation are different scoreboards

    A page can hold a top position on Google and still be absent from every AI answer. Measure both.

  2. 02

    Write the sentence you want quoted

    Models lift definitions, tables and sourced numbers. Put them where a model looks first.

  3. 03

    Off-site mentions do the convincing

    Assistants repeat what several trusted sources agree on. Your own site alone rarely gets you named.

The service behind this

AI SEO

See how we run this as a repeatable engagement, not a one-off.

View the service

Next case study

Reddit went from "risky channel" to their second-largest revenue source

D2C sleep products brand · India

Read it

Who led it

Prithvi runs this kind of work.

Prithvi Jadwani, Co-founder · Organic growth

Co-founder · Organic growth

Prithvi Jadwani

Runs the organic division: SEO, AI search visibility, Reddit and community marketing, and conversion. Six years of working out where buying decisions actually start.

Read more

What happens next

  1. Answer a few quick questions about your business. About two minutes.
  2. Pick a time that suits you for a free 30-minute call with a founder.
  3. We study your site before the call, so you leave with a plan, not a pitch.
Book your free strategy call

Free, no obligation. Your answers save as you go, so you can stop and pick up where you left off.