Old Mutual × PlainSpeak · Confidential

Optimising brands for
the AI era.

Answer engines are already deciding what people learn about Old Mutual. A proposal to measure that, and then shape it.

±56% of search volume ~60% zero-click Pilot: OMAI / AIIM

Prepared by PlainSpeakJuly 2026

00Agenda

What we'll cover.

The era is already here

How fast AI replaced search, and what that did to clicks

Why this matters for Old Mutual

YMYL scrutiny, PDF-heavy content, and answers given on your behalf

Changing the metrics

Why mentions replace clicks as the unit of performance

Measuring the black box

The prompt engine and the LLM listener

Adapting the content

Markdown summaries, PDFs, and technical metadata

Pilot scope and commercials

Three phases, and what it costs

01The era is already here

Four years, start to finish.

01Bottom line

In under four years, AI went from
curiosity to infrastructure.

~25%

of the world's population now uses an AI tool at least once a month.

Old Mutual × PlainSpeak · ConfidentialJuly 2026

01The era is already here

AI is rapidly replacing search.

±56%

of global search engine volume is now accounted for by AI tools

58%

of U.S. consumers use AI-powered search at least weekly

44%

say AI chat tools have changed how they look for information

~60%

of Google queries are zero-click — answered without visiting a site

Users instinctively reach for AI on complex topics — and professionals lean on it hardest, because it collapses the time cost of deep research. Financial services sits squarely in that zone.

02Why this matters

Financial services is a hard category.

Extra scrutiny applies

Financial services falls into Google's "Your Money or Your Life" category. AI engines hold YMYL content to a higher bar before they will repeat it.

But a citation is worth more

Harder to earn, more durable once earned. A mention in this category carries trust that a mention in a low-stakes category does not.

PDFs are a structural weakness

The sector runs on PDFs, and answer engines routinely struggle to decompose them. Vital content is effectively invisible.

This is already happening

AI engines are answering questions on Old Mutual's behalf right now — with or without your input into what they say.

02Are we already behind?

Good for specific prompts.

02Are we already behind?

Weaker for broader prompts.

Part two

How do we optimise
brands for AI?

Three moves: change the metrics, actually measure them, then adapt the content.

03We change our metrics

A mention is the unit now.

People use answer engines differently to search engines. The measurement has to start from that reality, not be retrofitted onto it.

What counts as a mention

Any answer-engine response that names your brand, quotes your content, or links to your site — usually all three at once.

Why it's detectable

When an engine mentions you it also pings your site with a crawler. That ping is the signal we can actually instrument.

03We change our metrics

Clicks are no longer king.

0.01%

Upper bound on citation CTR

0.001%

Lower bound

1 in 10k

To 1 in 100k mentions

Mentions are the new clicks

Clear, strong correlation between mention volume and conversions — most pronounced in B2B
CTRs falling in organic search too, even as impressions rise
Direct traffic is noticeably higher quality and more apt to convert
Effectively "dark social" on steroids: real influence, almost no attribution trail

If you keep scoring AI visibility on clicks, you will conclude that it isn't working — right up until a competitor owns the answer.

04We start actually measuring AI

Measurement is the challenge.

04We start actually measuring AI

We tackle it two ways.

We ask

We automatically query the major answer engines with a battery of prompts, and record the answers — and how they drift — over time.

We listen

We add a listener to your web infrastructure that detects when an engine mentions you, by catching the crawl it triggers.

Neither alone is sufficient. Asking tells you what engines say; listening tells you when they act. Together they close the loop.

04We start actually measuring AI

The prompt engine. We ask.

Asks 25 to 100 questions per day across the major engines
Records and compares responses over time
Tracks mentions — yours and your competitors'
Records when specific pages or downloads appear in, or vanish from, responses

PlainSpeak's Solution

We add your brands to our engine
We work with your team to craft the prompts
We report regularly on metrics and trends
We adjust and optimise prompts as the picture changes

04We start actually measuring AI

The LLM listener. We listen.

Runs at your server edge / CDN (Amazon CloudFront)
Serverless application — no infrastructure to maintain
Logs HTTP requests from on-demand crawlers to a database

PlainSpeak's Solution

We adjust the listener to Old Mutual's exact specifications
We work with your infrastructure team to install and validate
We build a live dashboard covering specific pages and the site as a whole
We analyse the data regularly and report on insights and trends

05We adapt and augment content

Make pages LLM-friendly.

LLMs love

Clear hierarchy
Technical and numerical detail
Text — plain HTML, markdown
Related links
Longer, comprehensive pages — cited ~1.6× more often than short ones
Pages that cite their own sources — cited ~2× more often than pages with none

LLMs hate

JavaScript — frameworks and menus
Data trapped in images and graphs
Important data in ephemeral elements like pop-ups
Ambiguous hierarchy

Does this mean rebuilding the site? No — roughly 90% of the benefit is available from simple, additive changes.

05We adapt and augment content

Same page. Two readings.

Markdown summary — what the engine sees

# Old Mutual Alternative Investments

## Private Equity Fund VI

### About the Fund

Old Mutual Private Equity Fund VI, established in
2023, focuses primarily on investments within South
Africa, with the flexibility to allocate up to 25% in
the Rest of Africa, with a preference for East Africa.

OMPE Fund VI had its first close in December 2023
and is open to new investment.

- [More About Old Mutual Limited](https://…)

The first 30 lines of the live HTML

<!DOCTYPE html><html><head><meta charSet="utf-8"/><meta
http-equiv="x-ua-compatible" content="ie=edge"/><meta
name="viewport" content="width=device-width, initial-scale=1,
shrink-to-fit=no"/><meta name="generator" content="Gatsby
4.24.8"/><meta data-react-helmet="true" property="title"
content="Old Mutual Private Equity Fund VI"/><meta
data-react-helmet="true" name="facebook-domain-verification"
content="5jrgqxr07ljo8aznnoc7pahxdqkud3"/><meta
data-react-helmet="true" property="og:type" content="article"/>
<span slot="hero-banner-header"><strong>Old Mutual</strong>
<strong><br /></strong><strong>Private Equity Fund VI
</strong></span><om-main-navigation slot="main-navigation"
primaryMenuWhite="false" homeUrl="/" hasSubMenu="true">
<om-main-navigation-menu activeMenuItem="ourfunds"><ul><li>
<a href="/our-business">Our Business</a></li>

05We adapt and augment content

Add markdown summaries to key pages.

10×

Fewer tokens to process

Tiny

File sizes

Clear

Explicit, unambiguous structure

Open

Widely supported, human-friendly

PlainSpeak's Solution

We decompose and summarise pages for clients, and assist in adding them to their sites.

05We adapt and augment content

Two more quick wins.

Markdown summaries of PDFs

PDFs are disproportionately common in financial services
They contain vital content
They are markedly harder for LLMs to read

PlainSpeak's Solution

We batch process PDFs for clients and create clean markdown versions, ready for upload.

Metadata and headers

Add an llms.txt
Add "alternate" tags in the document head
Tweak HTTP headers at the server

PlainSpeak's Solution

We create the technical and functional specifications, and work with your team to implement them.

06The virtuous circle

Every element feeds the next.

These pieces are not a menu. They compound — the measurement tells you what to change, and the changes give the measurement something to detect.

Observe

Listener and prompt engine produce insights.

Hypothesise

We collaborate on a batch of proposed changes.

Adjust

We augment content and infrastructure.

Repeat

Measure the delta. Feed it back in.

Old Mutual × PlainSpeak · ConfidentialJuly 2026

07Next steps

Start with a pilot.

Focused on Old Mutual Alternative Investments, and optionally AIIM.

Phase 1 · 3–4 weeks

Audit & baseline

Install the listener
Run the first prompt battery
Benchmark against competitors on 25 core queries

Phase 2 · 6–8 weeks

Technical implementation

Markdown summaries for priority pages
PDF batch processing
Crawler-serving infrastructure
llms.txt and header changes

Phase 3 · Continuous

Measurement & iteration

Monthly reporting
Prompt refinement
Content iteration driven by the data

What we need from Old Mutual

Infra/dev team mandate for listener installation
Sign-off on priority pages and PDFs for Phase 1
A stakeholder contact for ongoing prompt and content review

08Commercials

What it costs.

Setup — once-off

Component	Cost
Prompt tracking engine	R16,500
LLM listener (CloudFront)	R7,500
Dashboard	R4,500
Audit + baseline	Incl.
Content adaptation	±R27,500
Total setup	±R56,000

Ongoing — monthly

Component	Cost
Prompt tracking engine · 25 daily prompts	R1,350
LLM listener	R2,500
Dashboard	R1,750
Monthly insights & optimisations	R5,500

Total monthly	R11,100

Old Mutual × PlainSpeak · ConfidentialJuly 2026

Optimising brands forthe AI era.

What we'll cover.

Four years, start to finish.

In under four years, AI went fromcuriosity to infrastructure.

AI is rapidly replacing search.

Financial services is a hard category.

Extra scrutiny applies

But a citation is worth more

PDFs are a structural weakness

This is already happening

Good for specific prompts.

Weaker for broader prompts.

How do we optimisebrands for AI?

A mention is the unit now.

What counts as a mention

Why it's detectable

Clicks are no longer king.

Mentions are the new clicks

Measurement is the challenge.

We tackle it two ways.

We ask

We listen

The prompt engine. We ask.

The LLM listener. We listen.

Make pages LLM-friendly.

Same page. Two readings.

Add markdown summaries to key pages.

Two more quick wins.

Markdown summaries of PDFs

Metadata and headers

Every element feeds the next.

Start with a pilot.

What it costs.

Optimising brands for
the AI era.

In under four years, AI went from
curiosity to infrastructure.

How do we optimise
brands for AI?