Google changes the fields.
Attributes appear and disappear, categories get renamed and merged, new sections launch without announcement. A profile optimised eighteen months ago is missing fields that did not exist when it was built.
It decides whether you appear when somebody is ready to call. It feeds what AI says about you when they are still deciding. It changes every month, quietly, whether or not anyone is looking after it. Most businesses set it up once and have never touched it since.
0 of 0
ready-to-buy local searches returned a map box
eigenKOR, AI Answer Study, 22 August 20260%
of the research questions before that moment returned a generated AI answer
eigenKOR, AI Answer Study, 22 August 20260.0 v 0.0
star rating of the businesses Google shows against the ones it passes over
eigenKOR, Profile Attribute Study, 22 August 2026, replicated 30 August 2026In August 2026 we ran two hundred searches across Houston, Phoenix, Denver and Dallas, in seven local industries, collected in a single day from outside the cities being searched.1
Every single ready-to-buy search returned a map box. Seventy-two out of seventy-two. “Plumber near me.” “Emergency dentist.” “Urgent care open now.” Not most of them — all of them.1
Whatever else is changing about search, the moment your customer is ready to spend money they are handed three businesses and a map. Your profile is the entire basis on which Google decides whether one of them is you.
The same study found that the research happening before that moment now returns a generated AI answer 95.8% of the time.1 The profile feeds that too — the categories, the reviews, the photographs, the questions and answers. One record, two jobs, and almost nobody is maintaining it for either.
We asked the same fifty questions from Dallas and again from Fort Worth — two anchors about thirty miles apart inside one metro. The kind of answer Google gave barely moved: same question, same kind of result, 94% of the time.1
But of the twenty-nine questions where both cities returned a map box, only one produced the same businesses. Twenty-four produced a completely different line-up. Average overlap: 8%.1
Distance is not the whole story, though, and we tested that too. Across 800 businesses in ten categories and four metros, the ones Google showed sat a median 3.5 times closer to the city centre than the ones it passed over. But only 41.2% of them were among the three nearest.2 Six times in ten, Google reached past a closer business to show one further away. Something about those profiles earned it.
0%
of questions got the same kind of answer from Dallas and from Fort Worth, thirty miles apart
eigenKOR, AI Answer Study, 22 August 20260%
average overlap in the businesses those two map boxes named
eigenKOR, AI Answer Study, 22 August 20260 times
closer to the city centre, median, for the businesses Google shows
eigenKOR, Profile Attribute Study, 22 August 2026, replicated 30 August 20260%
of the businesses shown were among the three nearest
eigenKOR, Profile Attribute Study, 22 August 2026, replicated 30 August 2026This is why “get me to the top of the map” is not a coherent request, and why anyone who accepts it without qualification is either not measuring or not telling you. It is not one position. It is dozens of separate contests across your service area, and where you physically sit decides most of them before any work begins.
We measure that first and tell you which towns are winnable, which are contestable, and which are not worth a dollar. That conversation costs us work at the proposal stage. We would rather lose it there than at month four.
Most owners think of it as a listing — name, address, phone, hours. It is closer to a small structured website that Google hosts, and every field below is a live ranking and answer signal.
Primary category
Which searches you are eligible for at all. Get this wrong and no amount of downstream effort recovers it — you are competing in the wrong race.
Secondary categories
The additional races you may enter. Usually under-used, occasionally over-stuffed, rarely deliberate.
Services and products
Individually named, individually described. This is where the long tail of “do they do X?” gets answered.
Attributes
Wheelchair access, appointment required, women-owned, languages spoken, payment types. Worth completing so a customer filtering for one of them can find you — but we measured it twice, eight days apart, and got opposite answers: businesses Google shows carried fewer attributes on the first run and more on the second, which came out level once distance was held constant.2 A signal that flips is not a lever, and we will not sell you one.
The description
One of the few places you write your own words on a page you do not own.
Service area
Where you claim to work, which interacts with the proximity ceiling above.
Hours, including special hours
“Open now” is a filter applied constantly. Wrong holiday hours remove you on exactly the days demand spikes.
Photographs
Geotagged, captioned, refreshed. Customer-supplied images carry weight your own do not.
Video
Under-used and disproportionately cited — YouTube appeared in 31.9% of the AI answers we recorded, second only to Reddit and ahead of every other domain.1
Posts and offers
Freshness, and the only regularly-updated surface on the profile.
Questions and answers
A public question board attached to your business that most owners have never opened. Anyone can answer. Anyone does.
Reviews and your replies
Not the star average. We measured it: the businesses Google shows and the ones it passes over have the same rating — 4.9 against 4.9.2 What separates them is volume and substance. Systems that generate answers summarise what reviews say, so thirty that name the service, the town and the outcome beat a hundred reading “great job”.
We measured this rather than assert it. Eight hundred businesses across ten local categories in four metros, comparing the three Google shows against the seventeen it does not — same search, same city, all of them indexed and eligible.2
Then, because proximity could explain everything, we ran it again using only businesses within two miles of the centre — a hundred shown against four hundred and twenty-eight passed over, at comparable distance. Every gap that mattered got bigger.2
0.0 v 0.0
Reviews: median, within two miles, shown against passed over
eigenKOR, Profile Attribute Study, 22 August 2026, replicated 30 August 20260.0 v 0.0
Photographs: median, within two miles, shown against passed over
eigenKOR, Profile Attribute Study, 22 August 2026, replicated 30 August 20260 v 0
Categories: median, within two miles, shown against passed over
eigenKOR, Profile Attribute Study, 22 August 2026, replicated 30 August 20260% v 85.3%
Profile claimed: median, within two miles, shown against passed over
eigenKOR, Profile Attribute Study, 22 August 2026, replicated 30 August 20260% v 83.6%
Website attached: median, within two miles, shown against passed over
eigenKOR, Profile Attribute Study, 22 August 2026, replicated 30 August 20260.0 v 0.0
Star rating: median, within two miles, shown against passed over
eigenKOR, Profile Attribute Study, 22 August 2026, replicated 30 August 2026Roughly twice the reviews and twice the photographs of a comparable business standing the same distance away. Twice as many categories. And a star rating that is identical — the number every business watches, and the one measure in the study that separates nothing.2
Then we ran the entire study again eight days later — another eight hundred businesses, same design, same four cities. Every gap in that table survived it, and the review and category gaps widened.2 A single measurement is a claim; a measurement that repeats is a finding.
These figures describe the businesses Google shows. They do not prove that adding photographs causes a position — a business with more reviews and better photographs may simply be better run, and better-run businesses may rank for reasons no outside measurement can see. We publish the distinction because a firm that blurs it will sell you the wrong work.
A profile is not a thing you fix. It is a thing that decays.
Attributes appear and disappear, categories get renamed and merged, new sections launch without announcement. A profile optimised eighteen months ago is missing fields that did not exist when it was built.
The map box is relative. Holding position means matching what the businesses around you are adding — reviews, photos, services, answers — every month, indefinitely.
Recency is weighted. A hundred reviews from three years ago is not the asset it was, and the gap only closes by asking, consistently, forever.
Suggested edits from strangers, duplicate listings, wrong hours, a competitor marking you permanently closed. Unwatched, these stand.
New services, a new location, a price change, a member of staff who is now the person people ask for. None of it reaches the profile unless somebody puts it there.
The map grid across your service area, the profile scored field by field, your top three competitors on the searches that produce revenue, and the AI answers your category currently returns. You keep this whatever you decide next.
Town by town, winnable, contestable, or out of reach, with the reasoning shown.
Our own measurement found the businesses Google shows carry twice as many categories as comparable businesses it passes over — a median of two against one, and 60.0% using more than one against 48.6%.2 It is also the field most often left at whatever was chosen on the day the profile was created.
Attributes, description, service area, hours, products, opening dates.
Across every directory that carries your record, so the name, address and phone agree everywhere. We do not quote a number of directories in advance — the assessment establishes which ones actually carry you, and that count differs by business and by trade.
Existing questions answered properly, the obvious ones seeded and answered.
Matched to the profile, so the two corroborate rather than contradict.
Weekly
Continuous, triggered by job completion
Within one business day, every review, positive and negative
Monthly
Monthly, included
Weekly
Weekly
On change
Monthly, same grid, same queries
Quarterly
Monthly
We automate the repeatable work — publishing, requesting reviews, drafting responses, pushing changes out to directories and keeping them synchronised, and producing and distributing the monthly video. Doing that by hand is a waste of a person, and a business paying agency rates for copy-and-paste is being overcharged. What we do not automate is the decision to publish — see below.
What does not automate is the judgment: which primary category, which towns to contest, which questions your buyers actually ask in their own words, what the measurement means, and when the honest answer is that the profile is fine and the problem is somewhere else in the business.
Every published item is reviewed by a person before it goes out. Nothing reaches your profile unseen — not a post, not a review reply, not a photograph, not a video. The machine drafts and produces; a person approves and publishes. That is a deliberate cost we carry, and it is the difference between automation that saves your time and automation that speaks for you without supervision.
Nobody can guarantee a map position.
Proximity is weighted so heavily that some towns cannot be won from your address at any budget. We test it and tell you before you spend.
We will not promise a number of reviews.
We can promise the asking is consistent and the process is followed. Your customers decide the rest.
We do not write reviews, and we will not work with anyone who wants us to.
It is against Google's terms, it is detectable, and the penalty lands on your profile, not ours.
We will not tell you a profile field causes a ranking.
Our own measurement shows what the businesses Google shows have in common. It cannot show what caused it, and anyone presenting a correlation as a mechanism is either not measuring or hoping you will not ask.
We do not quote a price before we have measured you.
What this costs depends on the state of the record when we find it — how far the category has drifted, how many listings disagree, whether the reviews carry substance, whether there is any video at all. A firm that can price this off a phone call is selling a package, not a diagnosis.
If your customers do not search locally, this is not your lever and we will say so. If you are one of three businesses in your category within fifty miles, you may already win the map box without help — the assessment will tell you, and we would rather lose the sale than take money for work the geography is already doing.
Much of it, yes. The fields are free and the interface is public. What people run out of is not ability but consistency — the profile rewards someone doing modest work every week for years, and that is exactly the work that loses to a busy month. If you have someone who will genuinely own it, we will happily hand you the assessment and the priority order and let you run it.
Foundation corrections — category, services, listing consistency — can move measurable position within weeks. Review substance and photograph depth accumulate over months. We baseline everything before starting precisely so the difference between those two timescales is visible rather than argued about.
Each location is a separate profile, a separate map grid and a separate set of contests — the Dallas and Fort Worth result on this page is exactly that problem measured. Multi-location work is priced per location, because that is genuinely what it costs.
We respond to every review, and a considered public response to a bad one is worth more than the review costs you — future customers read the reply, and so do the systems summarising your reviews. Where a review breaks Google's policies we will flag it for removal, but flagging is a request, not a lever, and most stay up.
The corrections stay — categories, fields, listings and structured data do not revert. What stops is the accumulation: no new reviews being asked for, no fresh posts or photographs, nobody watching for suggested edits, and no measurement. Position decays slowly rather than collapsing, and it decays faster the more competitive the area.
Both, from one body of work. The categories, reviews, questions and photographs feeding your map position are the same record the AI answer layer reads. In our measurement the two surfaces were near-mutually exclusive — only 7% of searches showed both — so they are answering different questions, but they are drawing on the same underlying profile.1