A business appearing in an AI-generated answer is useful.

Being recommended may be better still.

But there is another question that receives much less attention:

Once AI recommends a business, does it give the customer a practical route towards that business?

That is the P — Pathway part of our RAMP Framework.

We decided to test it.

Using a realistic buying scenario involving a small skincare company looking for a packaging supplier, we ran the same search five times in Google AI Mode and recorded not only which businesses were recommended, but where the answer appeared to send the customer next.

Two findings stood out.

First, businesses with exactly the same recommendation frequency could receive very different pathways.

Second, a strong pathway could actually help expose a weakness in the recommendation itself.

Both suggest that simply counting AI appearances tells us only part of the story.

What Does Pathway Mean in the RAMP Framework?

The RAMP Framework looks at AI visibility through four connected questions:

R — Relevance: Was this a customer search where appearing actually mattered?
A — Appearance: Did the business actually appear?
M — Mention Quality: How was the business presented?
P — Pathway: Was there a practical route from the answer towards the business?

This experiment concentrated on the final part.

Imagine two companies are both recommended.

For the first, AI provides a direct link to a page explaining the exact service the customer needs, together with an obvious route to request a quote.

For the second, AI names the company but provides no obvious way to reach it.

Both could record one recommendation.

But are those two appearances commercially equivalent?

That is what we wanted to explore.

How We Tested It

We created a customer situation in an industry we had not previously used for our AI visibility experiments.

The prompt was:

I run a small skincare business in Manchester and need a UK packaging supplier for around 2,000 custom-printed product boxes at a time. I would prefer recyclable or FSC-certified packaging, relatively low minimum order quantities, and the ability to get samples before placing a larger order. Which three packaging suppliers should I consider, and why?

We then:

ran the identical prompt five times in Google AI Mode;
recorded the three businesses recommended in each response;
recorded the obvious website or citation pathway shown in the copied answer;
completed all five searches before investigating the suppliers;
and afterwards checked a selection of the official landing pages to see what happened when the customer followed the pathway.

This gave us 15 recommendation instances.

Because we were working from copied AI Mode responses, we have been careful not to assume that every interface element was necessarily preserved when the answers were copied.

Where no direct route was visible, we describe it as no obvious pathway in the copied response, rather than claiming that Google definitely displayed no pathway at all.

What Google Recommended

Across the five identical searches, Google AI Mode recommended ten different suppliers.

Supplier	Appearances
Custom Pack Studio	3
Vistaprint	3
Tiny Box Company	2
Lil Packaging	1
Emenac Packaging UK	1
Healey Packaging	1
The Product Boxes UK	1
Aly Packaging	1
Boxtopia	1
Pixartprinting UK	1

The supplier recommendations therefore varied considerably.

But for this experiment, our main interest was what happened after each recommendation.

Finding 1: Most Recommendations Had a Direct Official Pathway

In the copied responses:

13 of the 15 recommendation instances contained a clear direct link to the recommended supplier’s official website;
one appeared to provide an official-site route through a supporting citation rather than a prominent direct recommendation link;
one contained no obvious official pathway in the copied answer.

So approximately 87% of the recommendation instances we recorded included a clear direct official-business link.

That was higher than I expected.

And many of those links went beyond a generic homepage.

AI Mode surfaced pages relating specifically to subjects such as:

skincare packaging;
custom packaging;
low-MOQ packaging;
Manchester packaging;
product boxes;
and other areas closely connected to the customer’s requirements.

That suggests that Pathway should not necessarily be treated as a simple:

link / no link

measurement.

There can be meaningful differences in where the customer is sent.

Finding 2: The Same Appearance Score Can Hide Very Different Pathways

Custom Pack Studio and Vistaprint produced the clearest comparison.

Both appeared in three of the five searches.

If we measured only recommendation visibility, their results would look identical:

Custom Pack Studio: 3/5
Vistaprint: 3/5

But the Pathway results were different.

Custom Pack Studio

Custom Pack Studio appeared in Runs 1, 3 and 4.

Across those appearances, AI Mode provided direct official links each time.

Interestingly, it did not always select the same destination.

One response linked to a no-MOQ packaging page.

Another linked to a Manchester-specific packaging page.

Both related to the customer’s needs, but in different ways.

That suggests the route into a business’s website can itself be dynamic.

AI does not necessarily recommend a company and always send users to the same page.

Vistaprint

Vistaprint also appeared three times.

But its pathway varied much more.

In one run, AI Mode linked directly to a relevant Vistaprint packaging page.

In another, Vistaprint’s official site appeared among the surfaced sources but there was no similarly prominent direct link in the copied recommendation.

In the third, we could see no obvious Vistaprint pathway in the copied answer.

So across three consecutive appearances:

Appearance remained consistent.

Pathway did not.

That is an important distinction.

A visibility dashboard recording only that both businesses appeared three times could make their performance look equivalent.

RAMP gives us another question to ask:

What opportunity did each appearance actually give the customer to move towards the business?

Finding 3: A Strong Pathway Can Expose a Weakness in the Recommendation

The most interesting result came from Lil Packaging.

In the first run, AI Mode recommended Lil Packaging and provided a direct route to its official website.

From a Pathway perspective, that initially looked strong.

The customer could move directly from the recommendation to the supplier.

But remember one of the customer’s explicit requirements:

around 2,000 custom-printed boxes

When we subsequently checked the relevant Lil Packaging information, its website stated a minimum order quantity of 5,000 for customised printed packaging.

That appears difficult to reconcile with the customer’s requested volume of around 2,000.

And that creates an interesting RAMP result.

R — Relevance

The query was clearly commercially relevant.

The customer was actively looking for a supplier.

A — Appearance

Lil Packaging appeared as one of Google’s recommended choices.

M — Mention Quality

The suitability of the recommendation becomes questionable if an important published requirement conflicts with the customer’s requested order size.

P — Pathway

The Pathway itself was good.

Google gave the customer a direct route to Lil Packaging.

But following that strong Pathway helped expose the possible weakness in the recommendation.

This is one reason I don’t think the four parts of RAMP should be considered completely independently.

A strong Pathway does not automatically mean a strong recommendation.

Sometimes the Pathway helps the customer verify whether the recommendation should actually be acted upon.

This Suggests a Useful Refinement to “Pathway”

When I originally developed RAMP, the basic Pathway question was:

Was there a practical route from the answer towards the business?

I still think that works.

But this experiment suggests a slightly richer interpretation.

A useful Pathway question might be:

Did the AI answer give the customer a practical route towards checking the recommendation and taking the next relevant step?

That could include:

visiting the relevant service or product page;
checking prices or requirements;
requesting a quote;
obtaining a sample;
making a booking;
contacting the business;
or simply verifying whether the recommendation genuinely fits.

The Pathway is therefore not valuable merely because it generates a click.

It can help the customer continue the decision-making process.

A Strong Pathway Also Requires Somewhere Useful to Send the Customer

This experiment also highlighted something important from the business’s side.

AI can only surface a highly relevant official destination if the website contains a useful destination in the first place.

That does not mean:

create a page and AI will definitely surface it.

We cannot make that claim.

But consider what happened with some of the suppliers in this experiment.

AI Mode was able to send the customer directly towards pages concerning:

skincare packaging;
low-MOQ packaging;
custom packaging;
or packaging services specifically connected with Manchester.

Those pathways were possible because those destinations existed.

That raises a useful question for any business thinking about AI visibility:

If an AI system recommended us for this particular customer situation, do we actually have a page it could sensibly send that customer to next?

Suppose the answer is scattered across an entire website.

The business might mention specialist experience on one page, pricing on another, samples somewhere in an FAQ and the relevant service on a generic services page.

An AI system may still be able to piece some of that information together.

But from a Pathway perspective, a customer may be better served by a clear destination that brings the relevant information together.

That connects Pathway directly with the customer journey.

First identify the customer situation.

Then ask:

What would this customer reasonably need to see or do next?

A business cannot control whether an AI system will surface a particular page.

But it can control whether that useful page exists.

Pathway Starts Before the AI Recommendation

This may be one of the more practical lessons from the experiment.

Businesses often think about AI visibility as something happening entirely inside Google AI Mode, ChatGPT or another AI platform.

But part of Pathway begins on the website itself.

If a business wants customers looking for a particular service to be able to:

understand whether it is suitable;
see relevant evidence;
check important restrictions;
understand pricing;
request a sample;
obtain a quotation;
make a booking;
or contact the right person,

then those routes need to exist.

The AI system may choose to surface them.

It may choose a different page.

It may provide only the homepage.

Or it may provide no obvious direct route at all.

But the website still needs to give the AI somewhere useful it could potentially send the customer.

That is very different from saying that creating customer-focused pages guarantees AI visibility.

It doesn’t.

It simply creates better Pathway potential.

Sources Surfaced With the Answer and Customer Pathway Are Different Things

Another pattern appeared across the five runs.

Google AI Mode frequently displayed a mixture of sources alongside its answers.

Some were the recommended suppliers’ own websites.

Others were:

competing packaging companies;
industry websites;
social platforms;
or other third-party sources.

We should be careful about interpreting this.

We cannot see Google’s complete internal synthesis process, so the sources displayed alongside an answer should not automatically be treated as a complete record of everything used to construct it.

But we can observe something simpler.

The sources surfaced with the answer and the route given to the customer towards a recommended business were not always the same thing.

That gives us two useful things to record during AI visibility testing:

Sources surfaced with the answer

What websites has the AI interface shown alongside or in support of the response?

Customer Pathway

What practical route has the customer been given towards the business being recommended?

The distinction matters because a third-party source can appear alongside an answer while the customer is still directed straight to the recommended company’s official website.

Pathway Is Not the Same as Citation

This also explains why I would not treat citations and Pathways as interchangeable.

A citation tells us that a source has been surfaced in connection with the AI answer.

Pathway asks something more commercial:

What can the customer actually do next?

Sometimes a citation provides the pathway.

Sometimes AI supplies a prominent direct business link.

Sometimes it sends the customer directly to the relevant service page.

And sometimes a business can be recommended without an obvious direct route being preserved in the copied response.

Those are different outcomes.

What This Experiment Does — and Does Not — Show

This was a small exploratory experiment.

We tested:

one customer scenario;
in one industry;
on Google AI Mode;
five times;
producing 15 recommendation instances.

We should not assume that the same Pathway pattern will occur in every sector or on every AI platform.

Nor does this test establish why Google selected particular businesses or particular landing pages.

What it does show is that Pathway can reveal useful differences that recommendation frequency alone does not capture.

In this experiment:

most recommendations had clear official links;
some destinations were substantially more relevant than others;
the same business could receive different pathways across identical searches;
businesses with identical Appearance rates could have different Pathway experiences;
and a strong official pathway could help expose an apparent weakness in the recommendation itself.

Those are useful observations.

What This Adds to RAMP

The experiment started as a test of one letter in the RAMP Framework.

It ended up showing why the four letters work better together.

A business can:

appear for a highly relevant query;
receive a recommendation;
have a direct official website pathway;
and still be poorly matched to an important part of the customer’s requirement.

Equally, two businesses can achieve identical recommendation frequency while the customer pathway towards each is quite different.

So when testing AI visibility, it may be useful to move beyond:

Did we appear?

and ask:

Was the search relevant?

Did we appear?

How were we represented?

And what practical route did the customer receive from that recommendation towards our business?

Those four questions can produce a much richer picture than an appearance percentage on its own.

Final Thought

We began this experiment with a simple question:

AI recommended the business — but could the customer actually reach it?

In most of the 15 recommendation instances we recorded, the answer appeared to be yes.

But that turned out to be only the beginning of the story.

Some pathways were much more relevant than others.

The same business could receive different pathways across repeated searches.

Two businesses with identical recommendation frequency could have very different Pathway experiences.

And one strong official pathway helped reveal that the recommended supplier might not actually fit an important part of the customer’s requirement.

The experiment also highlighted something businesses can act on without making any claims about guaranteed AI visibility:

If you want AI to have the possibility of sending a customer somewhere genuinely useful, your website first needs to contain somewhere genuinely useful to send them.

That could be a specialist service page, a product page, a pricing guide, a booking page, a quotation form or another destination built around what the customer needs next.

Whether an AI system chooses to surface it is outside the business’s direct control.

Whether the useful destination exists is not.

And that may be one of the most practical implications of P — Pathway.

AI Recommended the Business — But Could the Customer Actually Reach It? Testing the “Pathway” in RAMP