How to Optimize When Local Customers Stay in Google Maps

Your local rankings look steady, yet calls and website sessions are falling. If those are the only actions in your report, the obvious conclusion is that local SEO has stopped working. That conclusion may be wrong.

A growing share of customers can evaluate a business, choose it and request directions without leaving Google Maps. Your job is no longer just to earn a listing that sends traffic elsewhere. You need to make the listing useful enough to complete the decision, support it with consistent evidence and measure what happens after the click disappears.

Diagnose a journey shift before declaring traffic lost

Corrected US portfolio data comparing Q1 2026 with Q1 2025 found that calls and website clicks each fell 15.8% while direction requests rose 31.3%. The same pattern continued in Q2, but at a slower rate: calls fell 11.9%, website clicks fell 12.5% and directions increased 21.1%.

The surface mix moved as well. In the US Q2 comparison, desktop Maps impressions rose 3.2% and mobile Maps impressions rose 30.4%, while mobile Search impressions fell 20.1%. That combination supports a practical working hypothesis: some local journeys are moving from search results into Maps, where customers can act without opening the business website.

It does not prove that every lost click became a store visit. These are portfolio-level changes, not a universal forecast for your locations. A direction request is a strong expression of intent, but it is not a confirmed arrival, purchase or booked appointment. Treat it as a distinct step in the journey and connect it to business outcomes wherever your systems allow.

• Discovery: Separate Search and Maps impressions, then split them by desktop, mobile, country and location.
• Decision: Report calls, website clicks and direction requests individually. A shift between them matters even when their combined total appears stable.
• Outcome: Compare those actions with bookings, qualified leads, online orders, store-level sales or another result the business can verify.
• Interpretation: If clicks decline while directions and downstream outcomes hold or grow, the journey may have migrated. If every action and outcome declines, investigate demand, visibility, listing quality and conversion instead of assuming a channel shift.

Rank tracking cannot settle the question. On 179 Google Business Profiles, AI-powered local packs often displayed two businesses rather than three, frequently omitted the call button and surfaced only 32% as many unique businesses as the traditional Map Pack. A tracker built around the traditional pack can therefore show a stable position while the customer sees a different set of choices.

When performance changes, inspect the actual Search, Maps and AI result experiences that matter to the location. Record whether the business appears, which competitors appear, what facts are shown and which actions are available. The visible interface is evidence your rank number cannot provide.

Do not apply a US benchmark blindly across countries. In the same Q2 comparison, EU desktop Maps impressions fell 34.7% while direction requests rose 13.1%. The smaller UK dataset moved differently again: mobile Maps impressions fell 70.8% while directions and website clicks increased. For an international brand, each country needs its own baseline and explanation.

Build a Maps listing that can finish the decision

Open your profile as if you have never heard of the business. Can you establish what it offers, whether it suits your need, when it is available, whether other customers trust it and how to reach it? Any unanswered question creates friction. It may also leave Google with too little confidence to answer that question on the business’s behalf.

Make the profile complete in decision order

1. Confirm identity. Verify the business name, primary category, address or service area, phone number and website destination. Multi-location brands should verify each location rather than assuming a central data feed is correct everywhere.
2. Confirm availability. Keep regular and special hours current. If a customer can book, reserve, order or request an appointment through a supported link, test that path from a signed-out customer view.
3. Describe the actual offer. Use the relevant categories, services, products and attributes available to the profile. Completeness means supplying useful facts, not adding promotional copy to every field.
4. Test every action. Call the listed number, open the website and booking links, and check where the directions pin ends. A correct-looking profile can still send a customer to a dead page, central switchboard or wrong entrance.
5. Assign ownership. Give one role responsibility for changes to hours, services, URLs, phone routing and location status. Profile accuracy deteriorates when each field belongs to a different team and no one owns the finished customer experience.

Completeness should be judged by whether a customer can decide, not by how many fields contain text. Remove stale offers. Avoid vague service descriptions. If two locations provide different services, represent the difference instead of copying one generic profile across the estate.

Align the profile, location page and entity markup

Local visibility now has two related layers. Traditional Maps rankings still depend on factors such as proximity, relevance, engagement and prominence. AI Mode and Gemini can layer web context, entity matching, brand authority and review sentiment onto the Google Business Profile. One layer influences whether the location appears as a map choice. The other influences whether an AI system has enough coherent evidence to recommend it or answer a specific question about it.

You cannot write your way around proximity. You can reduce uncertainty about relevance and identity. The profile, visible website copy and structured data should describe the same real business.

• Create a useful page for each location, with its real name, address or service area, phone number, hours, services and customer-facing destination links.
• Keep location distinctions visible in the page copy. A unique URL with generic text does not explain why that branch is relevant to a particular need.
• Use the most specific applicable LocalBusiness structured data to restate facts that are already visible on the page. JSON-LD should corroborate the page, not introduce claims a customer cannot see.
• Resolve conflicts between the profile, location page, schema, booking system and other business-controlled records. Do not choose a preferred version for reporting while leaving the public conflict in place.
• Write plain answers to recurring questions about services, suitability, access and other decision criteria the business can substantiate. Entity clarity comes from consistent facts in context, not repeated keywords.

Google Maps accuracy is especially important for Gemini because it can draw directly from Maps data. Do not mistake that connection for a complete cross-platform AI strategy. SOCi’s 2026 Local Visibility Index, a vendor benchmark rather than a universal census, found that the share of locations recommended was 1.2% on ChatGPT, 7.4% on Perplexity and 35.9% on Google. Profile accuracy averaged 68% on ChatGPT and Perplexity versus 100% on Gemini in that benchmark. The useful lesson is not that one percentage will predict your brand. It is that different answer engines can know different versions of the same location, so you must test them separately.

Turn reviews into answer-ready evidence

Reviews are no longer only a star rating beside your name. Their language can supply evidence about the questions a local customer asks before choosing: Was the place clean? Was it expensive? What was the atmosphere like? Did the business provide the particular service the customer needed?

Google now prompts reviewers with structured concepts such as atmosphere, price and cleanliness and encourages people to review places they have visited. Cleaner, more specific review data gives an answer system more material to summarize without sending the customer to a website.

Your review program should invite useful context without scripting praise or feeding customers keywords. A neutral request can ask the customer to mention the service or product they used and what mattered in their experience. That produces more decision value than a generic request for a five-star rating.

1. Ask after a real interaction. Make the request part of the customer handoff, receipt, completion message or other natural follow-up.
2. Keep the prompt neutral. Invite an honest description of the service used, the location and the factors that mattered. Do not tell the customer what sentiment or wording to publish.
3. Analyze themes by location. Separate repeated praise, repeated complaints, service mentions and unanswered questions. A multi-location average can hide a branch-specific problem.
4. Correct the underlying facts. If customers repeatedly misunderstand parking, pricing, appointment requirements or service availability, clarify the profile and location page where accurate. If the experience itself is wrong, fix operations before rewriting the description.
5. Respond for the next reader. Address the concrete issue, correct factual misunderstandings calmly and explain a resolved change when appropriate. Do not treat the response as a place to insert target queries.

Review quality may also affect whether a location enters an AI recommendation set. In the same vendor benchmark, locations recommended by ChatGPT averaged 4.3 stars and those recommended by Perplexity averaged 4.2. Those averages do not establish a rating cutoff, and they do not prove that raising a rating alone will earn a recommendation. They do show why reviews belong in AI visibility work alongside profile accuracy and on-site authority.

Measure the Maps journey all the way to a business outcome

A local dashboard should answer three separate questions: Were you visible, what action did the customer take and did the business receive value? Combining those stages into a single traffic chart conceals the very shift you need to understand.

Build a scorecard around the action mix

• Visibility: Search impressions, Maps impressions and observed inclusion in relevant traditional and AI-assisted local results, split by device and market.
• Profile actions: Calls, website clicks and direction requests shown separately as totals and as shares of all measured profile actions.
• Website behavior: Sessions and conversions from tagged profile links, including separate appointment, order or location-page destinations where available.
• Business outcomes: Qualified calls, completed bookings, orders, visits or store-level revenue. Use the outcome the business can measure consistently rather than claiming that every direction request became a customer.
• Data quality: Incorrect fields, unresolved profile-to-site conflicts, broken destinations and location pages missing decision-critical information.
• Review evidence: Rating, review volume and recurring themes by location, with operational issues separated from content gaps.

Do not add a call, a website click and a direction request together and label the total conversions. They represent different intentions and have different relationships to revenue. Keep the raw actions visible, then calculate downstream performance only where your systems provide defensible connections.

Run a repeatable local visibility cycle

1. Establish a comparable baseline. Preserve the device, surface, country and location splits. Use a comparable prior period when seasonality makes the immediately preceding period misleading.
2. Inspect the customer experience. Review the live profile, location page, action links, review themes, traditional local results and relevant AI answers. Capture what a customer can actually see.
3. Fix factual problems first. Correct identity conflicts, inaccurate hours, wrong categories, broken links and missing service information before rewriting copy or chasing more reviews.
4. Improve one evidence layer at a time where practical. A location-page update, profile cleanup and review campaign launched together may improve performance, but it will be harder to tell which gap mattered.
5. Read the whole journey. Compare changes in visibility, action mix and verified outcomes. A click decline with rising directions tells a different story from a decline across every stage.
6. Use outliers to choose the next action. In a multi-location account, investigate branches where action mix, review themes or downstream results diverge from similar locations. The portfolio average is a starting point, not a diagnosis.

This cycle also keeps paid and organic decisions grounded. Falling calls alone are not enough to prove that organic visibility failed or that paid search must replace it. You need to know whether customers disappeared, changed actions or finished the journey somewhere your report does not yet measure.

Key takeaways

• Google Maps can be the place where a local customer discovers, evaluates and chooses a business, not merely a route to the website.
• Stable traditional rankings do not guarantee stable exposure in AI-powered local results, and falling clicks do not prove that local demand has vanished.
• A complete Google Business Profile should answer decision questions and agree with the location page, structured data and customer-facing systems.
• Reviews provide answer-ready evidence about real customer concerns, but rating averages from a benchmark should not be treated as recommendation thresholds.
• Direction requests deserve equal visibility beside calls and website clicks, but they must not be reported as confirmed visits.
• Device, surface, country and location splits are essential because local behavior can move in different directions across markets.

In your next local report, place calls, website clicks and directions beside the business outcomes they are meant to produce. Then open each priority profile as a customer and remove the most consequential unanswered question. That is how you adapt to a local journey that may end in Maps without losing sight of the result that matters.

References

• Search Engine Land – Calls and clicks keep falling as Google Maps becomes the destination