Skip to content

Work

Visibility you can measure.

Answered Labs improves how businesses are understood, surfaced and recommended across search and AI-driven discovery. We track progress across recommendation presence, citation visibility, organic search, website discovery and the commercial actions that follow.

AI recommendation presence

+238%

Average increase in AI recommendation visibility

Businesses in this sample went from turning up occasionally in AI recommendations to turning up in most of the conversations that matter to them.

Recommendation visibility is not a ranking position. When someone asks an assistant for a shortlist, the assistant assembles an answer from whatever it can retrieve and trust at that moment, and the same question asked twice can produce two different lists. So we measure frequency instead of position: across a fixed set of prompts that mirror how real customers ask, how often does this business get named at all? An increase of this size usually means the business moved from being one of many plausible options to being one the model reaches for consistently.

What changed

The business became legible. Its services, service area and credibility were described the same way across its own site and the places the wider web talks about it, so there was less ambiguity for a model to resolve and more reason to include it.

How we measure it

A fixed prompt set is run on a schedule across several assistants. We record whether the business is named in each response, then compare the share of prompts that name it before and after.

Before 23%
After 78%

Before: 23 of 100 prompts After: 78 of 100 prompts

Citation presence

4.2x

Increase in AI citation presence

The sources sitting behind an AI answer changed, and the business's own pages started appearing among them.

Several assistants show their working by linking the pages they drew on. Those links are worth watching closely, because they show which corner of the web an assistant treats as reliable on a given question. Citation presence counts how often a company's own site, or a third-party page that describes it accurately, appears in that set. It tends to move before recommendation frequency does, which makes it a useful early signal that the underlying work is landing.

What changed

Pages that answered a specific question directly, in language a person would actually use, started being retrieved instead of pages that only described the company in general terms. A handful of accurate third-party listings helped as well.

How we measure it

For prompts where the assistant exposes its sources, we log every cited domain and page, then track how often the business or a page describing it appears.

Before 18%
After 76%

Before: 0.9 citations per 10 answers After: 3.8 citations per 10 answers

Search visibility

+167%

Average growth in non-branded discovery

More of the people finding these businesses had not heard of them beforehand.

Branded search is a measure of demand you already created. Non-branded search is a measure of whether you can be found by someone who only knows what they need. Someone typing the company name is already sold; someone typing what they need is choosing. Growth here is the part that compounds, and it matters twice over now, because conventional search results are still one of the main places assistants look when they assemble an answer.

What changed

Coverage improved for the questions customers actually ask, and the pages answering them were structured well enough to be understood without a visitor having to hunt for the answer.

How we measure it

Impressions and clicks for queries that do not contain the company name, taken from search console data and compared across matched periods.

Before 31%
After 83%

Before: 31% of baseline After: 83% of baseline

Commercial impact

+78%

Increase in qualified website actions from organic and AI discovery

The visits that arrived were more likely to turn into an enquiry, a booking or a call.

Visibility that produces nothing is a vanity metric, so this is the number we care about most. It counts the actions a business would be glad to receive: a form completed, a call started, an appointment booked. It rose faster than raw traffic did, which is the pattern you want. People arriving from a recommendation or from a question they typed themselves have usually already decided what they need, and they arrive further along than someone who clicked out of curiosity.

What changed

Traffic composition shifted toward higher intent, and the pages receiving that traffic made the next step obvious rather than making a visitor go looking for it.

How we measure it

Completions of the actions each business counts as commercially meaningful, segmented to organic and AI-referred sessions, compared across matched periods.

Before 39%
After 69%

Before: 39% of baseline After: 69% of baseline

What sits behind the number

There is no setting that makes a business rank inside ChatGPT. Anyone who tells you otherwise is selling something they have not tested. What actually moves is slower and less dramatic: a model can only recommend a business it can find, parse and have some reason to trust, and each of those is a separate problem.

So the work is rarely one thing. A change of the size shown above usually comes from several of these moving at once, and which ones matter varies by business:

  • technical site architecture, so pages can be crawled and read
  • content that answers the question a customer actually asked
  • entity clarity, meaning the web agrees on what this business is
  • search visibility, because assistants still lean on search results
  • authoritative third-party mentions and accurate directory listings
  • local signals, for businesses that serve a specific area
  • reviews and the public reputation attached to the name
  • citations from sources a model already treats as reliable
  • structured information that removes ambiguity
  • internal linking, page quality, relevance and authority

Worth saying plainly

No one can guarantee a fixed position in an AI answer because these systems are probabilistic and constantly changing. What we can do is materially improve the likelihood that a business is found, understood and recommended, then measure that progress across a consistent set of real customer questions.

How we measure

Six metrics we track, in this order

Each one answers a different question, and the last one is the only one that pays for the others.

01

AI recommendation presence

Does the business appear when customers ask AI systems who they should choose?

This is the headline question and the hardest one to fake. We run a fixed set of prompts written the way a real customer would ask, then record whether the business is named. Frequency across the set matters more than any single answer, because any single answer can be a fluke.

02

AI share of voice

Across a defined set of prompts, how often does a company appear relative to competitors?

Presence on its own can be flattering. Share of voice puts it in context by counting who else is being recommended for the same questions, which is usually the moment a client realises which competitors the model already trusts.

03

Citation presence

Which websites, profiles and third-party sources are being used to support the recommendation?

Where an assistant shows its sources, those sources are a map of what it currently treats as credible on that topic. Reading that map tells you which parts of the wider web are worth attention, and it often surprises people.

04

Search visibility

How well does the company appear across conventional search for relevant non-branded queries?

Conventional ranking has not stopped mattering. Several AI experiences retrieve from a live index, so search visibility feeds recommendation visibility, and it remains a large source of customers in its own right.

05

Entity strength

How consistently does the web describe what the company does, where it operates and why it is credible?

A business with three different addresses, two trading names and an unclear service list is expensive for a model to reason about. Consistency is unglamorous work and it is frequently the thing holding everything else back.

06

Commercial impact

Does improved visibility result in outcomes the business would actually be pleased to receive?

Qualified website visits, enquiries, bookings, calls, purchases and customers. If the first five measures move and this one does not, something in the chain is wrong and we would rather find that out than celebrate the chart.

How the work runs

A loop, not a launch

The useful part is the last step. Without it the first four are guesswork with an invoice attached.

01

Measure

Establish where the business actually appears today, across prompts and across search, before changing anything.

02

Diagnose

Work out which constraint is binding. Crawlability, clarity, content, reputation or authority.

03

Improve

Fix the constraint. This is usually technical and editorial work on the site itself.

04

Build authority

Earn the third-party signals that give a model a reason to trust the business.

05

Measure again

Re-run the same prompt set. Compare like for like, and be honest when something did not work.

Then back to the beginning

Related

The thinking underneath

Research

How we test the way assistants find, cite and recommend businesses.

Read How AI recommends local businesses

The signals that decide who gets named when someone asks for a local recommendation.

Read Pricing

What engagements cost and what is included at each level.

Read

Start with where you actually stand

The free audit shows you the prompts, who is being recommended instead of you, and what is getting in the way.

Get your free audit Book a call