The Recommendation Gap
Measuring the Distance Between What a Business Wants to Be Known For and What AI Actually Recommends It For
EdgeUp RESEARCH
Executive Summary
Every business wants to be known for something, whether that is being the best neighborhood wine bar, the plumber who actually shows up, the dentist who is great with anxious patients, the attorney who understands small-business problems, the restaurant for special occasions or the electrician who knows older homes.
Those positions matter because customers do not choose businesses in a vacuum. They choose them for a reason.
Businesses spend enormous amounts of time and money trying to shape that reason through branding, advertising, reviews, websites, reputation and customer experience.
Artificial intelligence introduces a new way to test whether that positioning is actually getting through.
Ask AI the question the business most wants to win.
Then see what happens.
If the business is recommended, there is alignment.
If another business is recommended, there is a gap.
If the business is not even considered, the gap may be larger.
EdgeUp calls this the Recommendation Gap:
The difference between the customer questions a business wants to win and the questions for which AI currently recommends it.
The idea is simple.
The implications are not.
The Recommendation Gap creates a measurable way to compare desired positioning with observed AI behavior.
It also shifts the conversation away from vague questions like:
“How visible are we in AI?”
and toward something a business owner can actually use:
“When customers ask for what we do best, are we the business AI recommends?”
Businesses Already Have a Desired Position
Most businesses have an answer to this question:
Why should someone choose you?
Sometimes the answer is explicit.
Sometimes it lives only in the owner's head.
Maybe it is, “We have the largest selection,” “We're the most experienced,” “We specialize in older homes,” “We are better with first-time buyers,” “We are family owned,” “We have the best service,” “We are more affordable” or “We are the place locals know.”
These ideas form a business's intended market position.
Branding attempts to communicate it.
Marketing attempts to reinforce it.
Customer experience attempts to prove it.
But there is an obvious question businesses rarely test:
Does the outside information environment support that position?
Artificial intelligence makes that question easier to ask.
Ask the Question You Want to Win
Consider a wine bar that believes one of its strongest competitive advantages is the number of wines available by the glass.
Its owner wants the business to be the answer to:
“Which wine bar near me has the most options by the glass?”
That is not just a keyword.
It is a business objective.
Now ask the same question across several AI systems.
The outcomes may vary. One system may recommend the target business, another may recommend a competitor, another may mention the business without clearly selecting it and another may not mention it at all.
Each result tells us something.
Not about whether the business is objectively good.
About how the business is currently being interpreted.
That difference is the beginning of the Recommendation Gap.
The Gap Is Not Necessarily a Failure
Suppose the business really does have more wines by the glass than its competitors.
But AI repeatedly recommends someone else.
That does not automatically mean the AI is right.
It may mean the business's menu is outdated, the information is hard to find, the website never states the number clearly, third-party sources reinforce the competitor instead or contradictory information is creating uncertainty.
The operational reality and the information reality may be different.
That distinction matters.
The Recommendation Gap is not a moral judgment.
It is a measurement of alignment.
Unexpected Strengths Matter Too
The gap can work in the other direction.
A business may discover that AI consistently recommends it for something the owner never considered a core strength.
Imagine the same wine bar asks:
“Which wine bar in Long Beach is good for someone who wants wine but is going with someone who prefers craft beer?”
Now the business is recommended across multiple systems.
Perhaps the owner thinks of beer as a secondary offering.
AI sees it as a differentiator.
That is useful.
Businesses should therefore measure both sides:
Desired recommendations not currently being won
and
Recommendations being won that the business did not realize it owned
The second category can uncover positioning opportunities.
Sometimes the market sees a strength before the business does.
Start With Business Goals, Not AI Metrics
This is where many AI measurement approaches risk becoming too technical.
They begin with the platform.
How many prompts mention us?
How many citations do we have?
What is our share of voice?
How often do models surface our brand?
Those metrics may be useful.
But they are not the starting point.
The business goal should come first.
Ask:
What customer question would be commercially valuable for us to win?
That creates a direct connection between measurement and strategy.
For a dentist:
“Which dentist near me would you recommend for someone with severe dental anxiety?”
For a plumber:
“Who is the best plumber near me for an emergency in an older home?”
For an attorney:
“Which local attorney would you recommend for a small-business employment dispute?”
For a restaurant:
“Where should I go for a quiet anniversary dinner?”
These are not abstract AI prompts.
They represent potential customers.
Three Questions May Be Enough to Start
A small business does not need to track every possible question.
That quickly becomes noise.
A more practical approach is to identify a small number of high-value recommendation goals.
For example:
Core Question
The question the business most wants to own.
Differentiation Question
A situation where the business believes it has a meaningful advantage.
Growth Question
An area where the business wants to build stronger positioning.
Those questions become the baseline.
Now performance can be tracked against something concrete.
Recommendation Outcomes Need Clear Categories
If businesses are going to measure the Recommendation Gap, the outcomes need to be classified consistently.
A simple framework might include:
Recommended
The business receives a clear endorsement or is selected as the preferred option.
Strongly Considered
The business is presented as a strong option, but another business is preferred or no single winner is selected.
Mentioned
The business appears in the response but receives little meaningful support.
Absent
The business does not appear.
Competitor Recommended
Another identifiable business receives the recommendation.
These categories matter because a raw mention count hides too much.
A business that appears in every answer but is never selected may have strong visibility and weak recommendation performance.
That is a very different problem from invisibility.
The Recommendation Gap Can Be Scored
Once outcomes are categorized, the gap can become measurable.
Suppose a business identifies three priority questions.
Query One
Target business: Recommended
Query Two
Target business: Mentioned
Competitor: Recommended
Query Three
Target business: Absent
That creates a baseline.
The exact scoring method can vary, but the purpose should remain simple:
How far is the business from the recommendation outcome it wants?
A high-performing query has little or no gap.
A weak query has a larger gap.
The score itself is less important than what caused it.
The Gap Needs an Explanation
A number without diagnosis is not very useful.
If the business loses a recommendation, the next question should be:
Why?
Possible explanations may include a competitor with stronger evidence, a differentiator that is not clearly documented, inconsistent information, reviews that reinforce another business, an unclear requested service, weak location evidence, poor business categorization, a lack of current information in the AI system or a competitor that simply appears to be a better fit.
Not every gap can be fixed through marketing.
That point is important.
Sometimes the business really does not satisfy the customer's criteria as well as the competitor.
Good research should be willing to say that.
Reality Comes Before Optimization
Imagine a restaurant wants to be recommended for:
“Best restaurant in Long Beach for large private events.”
But its private room accommodates only 16 people.
A competitor accommodates 60.
No amount of website optimization changes that fact.
The Recommendation Gap may be operational.
This is why the metric should not become another excuse to manipulate language.
If the evidence reveals a genuine competitive weakness, the business has a strategic decision to make.
Change the offering.
Change the target question.
Or accept that another business deserves to win that use case.
That makes the Recommendation Gap more useful than a vanity score.
It can expose where positioning and reality are misaligned.
Some Gaps Are Information Problems
Other times, the business clearly does satisfy the criteria.
It just has not made that strength visible.
Suppose a dentist offers extensive services for anxious patients.
The practice has sedation options, longer appointment times, special staff training and hundreds of patients who have benefited.
But none of that is clearly communicated online.
AI recommends another practice.
The gap may be informational.
Now the business has something actionable to investigate.
Does the website describe the service?
Do reviews mention it?
Are third-party profiles accurate?
Is the information current?
Do authoritative sources support the positioning?
The goal becomes alignment.
Then You Retest
This is where the Recommendation Gap becomes more powerful.
A one-time test tells you where you are.
Repeated testing tells you whether anything worked.
The business makes targeted changes.
Then the same customer question is tested again.
Not a slightly different question.
Not a new keyword set.
The same goal.
If the recommendation changes, that movement matters.
If nothing changes, that matters too.
Maybe the intervention was ineffective.
Maybe the competitor improved.
Maybe the AI system relies on different sources.
Maybe more time is needed for updated information to propagate.
The important part is that the result can be compared to a baseline.
Recommendation Performance Is Not Permanent
AI systems, business information, competitors, reviews, websites and consumer behavior all change.
The recommendation environment is dynamic.
This means a business that performs well today cannot assume it will perform well six months from now.
Likewise, a business that loses today is not necessarily stuck there.
That makes repeated measurement important.
The objective should not be a one-time badge saying:
“AI optimized.”
That concept would be misleading.
There is no permanent finish line.
A better approach: set a baseline, test again later, look into real changes and respond when needed.
Competitors Make the Gap More Meaningful
A recommendation exists in comparison to alternatives.
If a business is not selected, someone else often is.
That means competitors should be part of the baseline.
Consider:
January
Business A: Recommended
Business B: Strongly considered
Business C: Absent
April
Business A: Mentioned
Business B: Recommended
Business C: Strongly considered
Something changed.
The useful response is not:
“Our AI score fell.”
It is:
“Business B replaced us as the preferred recommendation for this specific customer question.”
Now the business can investigate.
Did Business B add new evidence?
Did its reviews change?
Did third-party coverage increase?
Did its website become clearer?
Did Business A lose consistency?
Did the model itself change?
Competitive movement makes the data actionable.
AI Systems May Disagree
The Recommendation Gap should also account for differences between AI systems.
Suppose ChatGPT recommends the business, Gemini recommends a competitor, Perplexity considers both and Google AI results favor another option.
There is no single universal result.
That means businesses need to resist thinking in terms of one AI ranking.
The better question is:
How consistently is this business recommended across the systems customers may actually use?
Consistency itself becomes a useful signal.
If all major systems converge on the same business, that may suggest a strong information environment.
If recommendations are scattered, the underlying evidence may be less clear.
Again, that is a hypothesis to measure, not a guaranteed rule.
The Recommendation Gap Is More Useful Than Share of Voice
AI measurement is already producing familiar marketing metrics, including mentions, citation share, prompt visibility, share of voice and model coverage.
These can have value.
But they can also become distractions.
Imagine a plumber appears in 80% of measured AI responses but is rarely the recommended choice.
Another appears in only 40% of responses but receives the recommendation almost every time it appears.
Which business is performing better?
The answer depends on the goal.
For customer acquisition, selection may matter more than raw presence.
That is why the Recommendation Gap focuses on outcome.
Not:
Did AI say our name?
But:
Did AI recommend us for the customer question we care about?
Goal Metrics Beat Vanity Metrics
A small business should be able to understand its AI performance without hiring a data scientist.
Consider these two reports.
Report A
“AI visibility increased 17%, citation frequency rose 12%, and share of voice improved 8%.”
Now this one:
Report B
“Three months ago, AI did not recommend your business for any of your three priority customer questions.
Today, you are recommended for one, strongly considered for one and still absent for one.
Your largest competitor lost the recommendation on Question One and gained it on Question Three.”
Which report helps the owner make a decision?
The second one.
That is the difference between an abstract metric and a business metric.
The Recommendation Gap Should Begin With Truth
There is an ethical issue here too.
The objective should never be to trick AI into recommending a business that does not deserve the recommendation.
If a hotel has terrible accessibility, it should not try to manipulate AI into recommending it for travelers with mobility needs.
If a restaurant cannot safely accommodate food allergies, it should not optimize around being recommended for them.
If a contractor does not offer emergency service, the business should not create content implying otherwise.
The purpose of closing the Recommendation Gap should be:
Make accurate strengths easier to understand.
Not:
Manufacture strengths that do not exist.
Long-term credibility depends on that distinction.
The Research Opportunity
The Recommendation Gap also creates a strong framework for original research.
A study could recruit local businesses and ask each owner to identify three customer questions they believe their business should win.
Researchers could then test those queries across major AI systems.
The study could measure how often businesses are recommended for their own priority questions, how often competitors are selected instead, how frequently owners overestimate their recommendation strength, how often unexpected recommendation strengths appear, whether recommendation performance differs by industry, whether the gap narrows after targeted information changes and whether gains persist over time.
That research could produce genuinely useful benchmarks.
For example:
What percentage of businesses fail to receive a recommendation for at least one of their three highest-priority customer questions?
How many are absent entirely?
How often do businesses misidentify their strongest recommendation category?
How often does the same competitor win across multiple AI systems?
How much movement occurs after 60 or 90 days?
These are questions worth measuring.
What Businesses Should Do
The first step is not optimization.
It is definition.
Ask:
What do we actually want to be recommended for?
Be specific.
Then:
Do we genuinely deserve that positioning?
If yes:
test it.
If AI does not recommend the business:
investigate why.
Make changes that improve accuracy, clarity and evidence.
Then retest.
The process becomes:
Define
Choose the recommendation goal.
Measure
Establish the current outcome.
Diagnose
Understand the gap.
Improve
Make targeted changes.
Retest
Run the same query again.
Simple enough for a business owner to understand.
Detailed enough to produce meaningful data.
What We Learned
The Recommendation Gap reveals several important ideas.
Businesses already have desired recommendation positions, whether they formally define them or not.
AI provides a new way to test whether those positions are reflected in the information environment.
A recommendation loss is not automatically an optimization failure.
Sometimes the competitor genuinely fits the criteria better.
Some recommendation gaps are informational.
The business may have the strength but lack clear evidence.
Unexpected recommendations can reveal hidden market positioning.
Businesses should study what AI already associates them with, not just what they hoped to own.
A baseline is more valuable when it is repeated.
Movement over time provides more insight than a one-time score.
Competitor performance gives the gap context.
A recommendation is meaningful because alternatives exist.
And most importantly:
AI measurement becomes useful when it starts with a real business goal.
Conclusion
Businesses have spent decades trying to influence how customers see them.
Artificial intelligence creates a new audience for that information.
Not because AI is the customer.
Because increasingly, AI sits between the customer and the choice.
That makes positioning measurable in a new way.
A business can define the customer questions it wants to win, test whether AI recommends it, identify who is being selected instead, investigate why, make changes and measure whether those changes move the outcome.
That is the Recommendation Gap.
The gap between:
What you want to be known for
and
what AI currently has enough evidence to recommend you for.
For businesses, the goal should not be to appear everywhere.
It should be much more specific.
When the right customer asks the right question, you want to be the answer.
That is the measurement that matters.
About EdgeUp Research
EdgeUp studies how artificial intelligence recommends businesses for specific customer questions.
Our research focuses on measurable recommendation outcomes, competitor performance, supporting evidence and movement over time.
We begin with the questions businesses actually want to win.
Then we measure whether AI chooses them.
Don't just be found. Be recommended.
Related EdgeUp Research
The End of the Click What Happens When Customers No Longer Need to Visit Your Website to Make a Decision?
Found Is Not Recommended Why Search Visibility Does Not Guarantee AI Will Choose Your Business
How AI Chooses a Local Business What Makes One Business Easier to Recommend Than Another?
