Your map-pack visibility, your calls, or your website clicks dropped, and someone told you AI Overviews took them. Maybe. The first thing we do is find out whether that is actually what happened, because the fix for an AI Overview problem and the fix for a suspended profile look nothing alike. Sometimes the answer is a correction or an appeal rather than a recovery engagement, and you will hear that from us before anyone proposes paid work.
Below is the order we work a visibility drop, and the places where that order ends the conversation instead of starting an engagement.
The first campaign we talk people out of
Someone has told you distance stopped mattering, so the plan is authority work aimed at neighborhoods across town. We will not quote that campaign on the strength of that claim. First we test whether the opportunity is really there in your market, and the study being quoted at you is the reason why.
The number everyone repeats is real. Local Falcon’s May 2025 study ran roughly 60,000 simulated searches across 4,423 businesses in 20 countries and found almost no relationship between distance and rank position inside an AI Overview: a correlation of 0.001. What gets dropped is the clause attached to it. That describes rank order once a business appears at all. On appearing, the same study found a smaller difference: 72.0% within a mile of the searcher, 68.5% between one and two miles.
Three and a half points. Enough to suggest proximity still bears on whether you enter the set; nowhere near enough to call it a gate.
That is a narrower finding than either sales pitch built on it. It does not establish a radius, and it does not tell us whether an East Nashville business can compete in Green Hills. It tells us that once you are in the room, being closest is not what wins.
So we test that question in your market instead of converting a twenty-country study into a Nashville radius rule. And once proximity stops explaining the order, we inspect what the result can actually draw from: profile relevance, site content, third-party references, reviews, how consistently the business reads as one entity across the web. That May 2025 study did not weigh those signals against each other, so we will not pretend it did.
Before we talk about AI, we check whether AI is your problem
A visibility drop has a cause. We check the older failure modes before assigning it to the newest SERP feature. If your organic rankings fell too, and not just the map pack, the wider diagnostic starts here: algorithm timing, technical faults, content decay, competitors. This page takes the branch where the pack went quiet. Our free audit exists to find that cause: where you currently rank and what put you there, keyword gaps, technical faults, what competitors did, and where the site loses the people who arrive. Applied to a vanished map pack, the sequence runs:
- Is the profile suspended or restricted? Nothing else matters until this is answered, and it has nothing to do with AI.
- Does the profile still lead somewhere that works? We check the website link, the page it lands on, and whether that page is reachable at all. If your organic visibility fell alongside the pack, we branch into the technical diagnostic. A crawl or indexing fault explains a website loss. It does not, by itself, explain why the profile stopped appearing.
- Do the categories actually describe the business? We pick the most specific accurate primary category, then only the additional ones that represent real parts of the business. Not one per service: individual jobs like furnace repair and AC installation belong in your service list and your site content, not in the category field. A category that misdescribes you can weaken relevance. It does not let us diagnose every missing query as a category fault.
- Is the business information materially consistent wherever it appears? A wrong phone number, an old address, a duplicate listing, a business name that does not match: those create a real identity problem. “St” against “Street” is a postal variation, not contradictory information, and we decline to bill both as if they were the same job.
- Did the query mix change, or did you? We include this step because its answer can show that no new engagement is warranted. We ask it last, and we mean it.
When it turns out to be a suspension or a miscategorization, that is what you hear, and what follows is a small piece of work or none at all.
Which queries you lost decides whether we take the job
Before we quote anything we want to know which searches went quiet, because the answer decides whether this is our problem to solve or somebody else’s story about it.
AI Overviews do not turn up evenly, which is the part “AI took your traffic” leaves out. Whitespark’s May 2025 study, 540 manual queries across six verticals in three US cities, found them on 15% of local-intent searches, the “bakery Nashville” kind, and on 92% of informational and 97% of hybrid ones, the “average cost of furnace replacement in Nashville” kind. Local packs ran on 93% of those local-intent queries. Local Falcon’s differently constructed 2025 query set logged 40.2% overall. The two numbers are not directly comparable: different samples, geographies, devices, query classes. What both show is that query intent strongly shapes which format appears, though the exact ratios move with category, market, device, and study design.
So the pack is not disappearing uniformly. The format splits by intent. “Emergency plumber near me” is still far more likely to return a pack; “how much does furnace replacement cost” is far more likely to return an Overview. What those percentages describe is which surface appeared, not how much traffic or how many leads each one produced, so we measure the traffic and the actions separately rather than reading a presence rate as a traffic share.
Underneath that sits the number we actually plan against. Sterling Sky tracked AI local packs surfacing 5,943 unique businesses across a query set where traditional three-packs surfaced 18,330: roughly a third of the inventory. AI packs also tend to show one or two businesses instead of three, and they drop the call button.
That is the threat we are pricing against. It is not only that you might rank lower. The list itself got shorter, which means a position that used to keep you visible can now sit under the cutoff.
Which is why we will not diagnose this from Search Console alone. Search Console shows which website queries and pages moved. It does not tell you whether an impression came from an AI Overview, a pack, or a plain organic result, and the profile’s own actions, the calls and the direction requests, never reach it at all. So we read it next to Google Business Profile performance, next to tagged profile links in GA4, and next to current samples of the actual SERPs your customers see.
When the losses cluster in informational searches and those SERPs now carry AI Overviews, the AI explanation becomes plausible and we can scope entity and content work against it. When “near me” visibility fell while the SERP format held steady, something else broke, and selling you an AI recovery package would be taking your money under a fashionable headline. We do not sell services a business does not need, which sometimes means telling you the drop is not the thing you were told it was.
What we correct inside the profile, and whether anything ongoing is warranted
The profile is where we start on a visibility drop, and not because it is easy. It gives us the shortest path to correcting data the business controls directly, rather than waiting to be re-evaluated against everyone else’s. That is not a promise of a fast ranking response.
Profile work is scoped to the problem we find, not sold as a fixed shape. Setup and optimization covers what is broken or blank: categories, service area, the fields nobody filled since the listing was claimed, the hours that send someone to a locked door. Review and photo work, and anything recurring, we propose only when the diagnosis gives it a job. They are not automatic steps after a drop, and selling a year of scheduled posting to paper over a category error is the easier invoice and the wrong one.
One correction we owe you here, because our own package sheet still says otherwise. Google discontinued the My Business Q&A API on November 3, 2025, and in December described the change in its own words: customers now ask in Google Maps and get an instant answer drawn from your answers and relevant reviews, rather than scrolling old threads. What closed is the legacy format: posting and managing new public Q&A threads is no longer a current deliverable, so we have taken it out of this offer.
What replaces it is Ask Maps, and we will keep the claim inside what is verified. As of March 2026 it was live only in the US and India, only on mobile, and it does not appear for every category. Its answers pull from your profile, your reviews, and your website, which means the way to influence it is the work we already do rather than a new line item. Other agencies may build real services around it, answer-source monitoring and the like, and that is a legitimate thing to sell. We have not built one, and we are not relabeling profile, review, or website work as a new Q&A service to fill the gap.
One line holds no matter what the drop looks like. A review request cannot be routed around the customers you expect to complain. Google’s policy forbids soliciting positive reviews selectively, and the consequences it names run from reviews being removed to restrictions landing on the profile itself. Compliance, not tactics.
The neighborhood page test
One Nashville is a fiction, and our practice is built on that fact. It is also the premise most abused by agencies selling location pages.
A neighborhood page needs a reason to exist that survives being said out loud. Show us the work you do there, the intake or client records tied to that area, Search Console queries that carry the neighborhood name in them, grid-based visibility data for that patch of the city, or a door you actually keep there. Search Console will not tell you a searcher stood in East Nashville; it has no dimension for that below the country level. It will tell you someone typed “East Nashville,” which is a different fact and still a useful one. Absent that, the honest recommendation is not to build it. Swap the place name and change nothing else and you are inside the territory Google’s spam policies mark as doorway abuse: a run of substantially similar local pages, built for similar searches, all pushing the reader to the same place and adding nothing local en route.
On one point we are inflexible, and it costs nothing to comply with: use the name residents use. East Nashville. The Gulch. 12 South. Not Nashville East, not Nashville Gulch, not Twelfth South. It matches the language your customers actually type, it keeps the page unambiguous, and it stops you inventing naming variants your search or customer data does not support. We are not claiming one variant decides whether Google credits you with serving the area.
We do not set neighborhood-density targets, because there is no such thing to hit. Google has been dismissing the idea of an ideal keyword density since Matt Cutts made a video about it in 2011, and John Mueller was still answering the question with a flat “no” a decade later. An agency quoting you a percentage is selling you a number Google has spent fifteen years saying it does not use.
Timing: our range, and why it is not a promise about this
We will not give you a date, and the reason is that nobody has one to give. Google publishes no fixed or guaranteed recovery timetable. Its own guidance says crawling can take from a few days to a few weeks, and that requesting a crawl does not guarantee inclusion in results instantly or at all. We are not turning broad crawl guidance into a map-pack recovery date, and an agency that does so to make a 48-hour package sound urgent is selling you certainty it invented.
Here is our own published range instead. Across the businesses we work with, measurable movement in search generally shows up somewhere in the 60 to 90 day window, and the fuller picture takes four to six months, with the spread depending on how contested the category is, what shape the site arrived in, and how much had to be rebuilt. Those figures describe our client base generally. They are not an AI Overview recovery benchmark. No published recovery case of ours exists that would earn that framing. A number borrowed from a study is not our recovery benchmark, and we will not present one as though it were.
What can be corrected immediately is the profile data itself: an inaccurate category, wrong hours, a dead link. Each is a different failure. The category problem is the relevance one above; wrong hours misstate when you are open; a dead link breaks the path to the call. Then we measure whether visibility or actions move, rather than assigning the correction a recovery date in advance.
What we would rather you spent the money on
Our honest recommendation, at the point where it costs us money: if the AI list holds one or two names instead of three, the demand you should be protecting is branded search. It depends least on discovery, because the searcher already knows your name and is not asking Google to find them a plumber. Not immune to AI features: in that same 2025 dataset, Local Falcon logged Overviews on 10.5% of navigational queries. But it is the demand least exposed to a shrinking list.
You build that demand the slow way: chamber membership, a neighborhood sponsorship, coverage in a Nashville publication, work good enough that your name comes up in a Facebook group. Track it in Search Console by filtering for queries containing your business name, and watch the absolute impressions and clicks, not the branded-to-generic ratio: the ratio climbs on its own when generic falls, which tells you nothing about whether anyone learned your name. Where the profile is part of the path, read its performance data alongside.
The profile work we can scope today. The listings and local links sit inside a broader engagement, and the audit will tell you whether they belong in yours. The rest of what builds a name in Nashville is not something we can invoice, and we would rather say that than pretend the part we sell is the whole answer.
Request the audit
Ask us for the free audit. For a map-pack drop it runs the sequence above: whether the profile is suspended, whether it still leads somewhere that works, whether the categories describe the business, whether your business information is consistent, and whether the mix of queries you lost actually moved. Then we read the current SERPs to see whether AI Overviews are even a plausible contributor. Several of those answers end in a correction or a conclusion rather than a proposal, and we would rather hand you one now than bury it inside a retainer.
The review is Nick Rizkalla’s own work, not a junior’s checklist. Whatever engagement comes after it, if any does, is billed month to month.