Google shipped a new model into AI Mode on August 14, 2026, one day after releasing Gemini 3.7 Flash to developers. Search Engine Journal reported that the update replaces Gemini 2.5 Flash as the selectable model inside AI Mode's "Ask anything" bar. For the average search user, that sounds like a backend tweak. For a trades business that depends on Google traffic, it is the third model swap inside AI Mode this year, and each one has moved the citation floor.
What Gemini 3.7 Flash changes under the hood
Google described Gemini 3.7 Flash as its "most intelligent workhorse model yet for coding and agents" when it launched on August 13. The model brings better multi-step planning, improved tool use, and stronger instruction following than its predecessor, Gemini 2.5 Flash. According to remio.ai's analysis, the model entered AI Mode less than 24 hours after its public release, which means Google is no longer waiting for major redesign events to upgrade Search. It is shipping model improvements into AI Mode as they arrive.
For local businesses, the practical question is whether a smarter model recommends different businesses. The answer based on the pattern from prior swaps is yes. When Google moved from Gemini 2 to Gemini 3 earlier this year, the number of cited sources per AI answer increased, freshness carried more weight, and entity-rich websites gained share at the expense of thinner ones. Some businesses moved up in AI answers. Others dropped out entirely. The same dynamic is likely with 3.7 Flash, because the model is better at synthesizing multi-step reasoning, which means it can compare businesses across more dimensions before picking which ones to name.
Why this swap matters for plumbers, HVAC, and electricians
Gemini has the highest recommendation rate of the three major AI platforms for local businesses. SOCi's 2026 Local Visibility Index, which analyzed over 350,000 business locations across 2,751 brands, found that Gemini recommends 11% of local business locations. ChatGPT recommends 1.2%. Perplexity recommends 7.4%. All three are dramatically lower than the 35.9% appearance rate in Google's traditional local 3-pack.
Search Engine Land covered the report and noted the key difference: AI systems prioritize confidence and risk reduction over breadth. A business with middling Google reviews can still rank in the local 3-pack based on proximity. AI Mode skips it. SOCi found that businesses recommended by ChatGPT averaged 4.3 stars, compared with 3.9 on Gemini and 4.1 on Perplexity. The AI models are filtering for quality signals that traditional local search does not enforce as strictly.
For trades businesses, this cuts two ways. A well-reviewed plumber with a complete Google Business Profile and consistent data across directories is more likely to benefit from a smarter model, because Gemini 3.7 Flash can better match the business to the query intent. A plumber with incomplete hours, missing service descriptions, or inconsistent name-and-address data across platforms is more likely to get filtered out, because the model has more dimensions to compare and the business fails on more of them.
What changed between 2.5 Flash and 3.7 Flash for local search
Google has not published a changelog specific to local search behavior inside AI Mode. Based on the model's documented capabilities and the reporting from Search Engine Journal and remio.ai, three things are different:
1. Better multi-step reasoning. Gemini 3.7 Flash is designed for more diligent multi-step planning, which means it can chain together a user's question across more steps before producing an answer. A query like "my heater stopped working, who can come today in Bergen County" requires the model to understand urgency, service type, geography, and availability. A model that handles multi-step reasoning better is more likely to pick a business whose Google Business Profile and website explicitly state emergency service availability, same-day response, and the service area.
2. Improved tool use. The model is better at calling external tools and APIs during its reasoning, which in AI Mode means it can query Google's local index more effectively. This benefits businesses with richer structured data on their profiles and websites. LocalBusiness schema that clearly defines service type, area served, hours, and contact information gives the model more to work with.
3. Greater instruction fidelity. Gemini 3.7 Flash follows instructions with higher fidelity than 2.5 Flash, which means it is less likely to hallucinate a business that does not match the query. For trades businesses, this reduces the chance of showing up for the wrong query but also raises the bar for showing up at all. Your profile needs to match what people actually ask.
How to check if the swap affected your business
The fastest way to tell whether Gemini 3.7 Flash changed your AI visibility is to run the same queries a homeowner would run and compare the results to what you saw a month ago.
Open Google AI Mode and search for your trade in your service area. Use natural language, not keyword strings. "Best plumber in Hackensack" is fine. "My sink is leaking, who can come today near me" is better. Do this across five to ten queries that reflect how customers actually talk.
Check whether your business appears in the AI answer. Note whether you appear in the cited sources at the bottom of the AI response. Then compare to what you saw in July, before the 3.7 Flash swap.
If your business dropped out of AI answers after August 14, the model swap is the likely cause. Grounding drift is the documented pattern where AI search results shift between queries even when your website has not changed. A model swap amplifies it, because the new model applies different weights to the same data.
What to do right now
Update your Google Business Profile. Gemini pulls directly from Google's local index, and SOCi found that business profile accuracy on Gemini is 100% because it is grounded in Google Maps data. Your GBP is the single most important input the model has. Fill in every field you can: service areas, hours, service descriptions, photos, and the Q&A section. A complete profile gives Gemini 3.7 Flash more to cite.
Add or update your service descriptions. The model's improved instruction fidelity means it matches businesses to queries more precisely. If your GBP service description says "plumbing repairs" but a homeowner asks about "frozen pipe repair," the model may not connect the two. Write service descriptions in the language your customers use.
Clean up your directory listings. ChatGPT pulls from Bing, and Perplexity pulls from multiple sources. Consistent name, address, and phone across Google, Bing, Yelp, Better Business Bureau, and industry directories signals that your business is established and active. Inconsistent data does the opposite.
Publish fresh content. Prior model swaps increased the weight on content freshness. Gemini 3.7 Flash is better at multi-step reasoning, which means it can evaluate not just whether your business exists but whether it is active. A blog post about a recent service call, a seasonal maintenance tip, or a local code change signals activity in ways the model picks up.
Run a free AI visibility audit. FirstCall Visibility checks how your business shows up across ChatGPT, Gemini, and Perplexity for the queries your customers actually use. Run your free audit here.
FAQ
What is Gemini 3.7 Flash?
Gemini 3.7 Flash is Google's latest AI model, released August 13, 2026. Google described it as its most capable workhorse model for coding and agent tasks. It replaced Gemini 2.5 Flash inside Google AI Mode the next day.
How does Gemini 3.7 Flash affect my trades business?
The model powers Google AI Mode, which generates AI answers when users search for local services. A model swap can change which businesses appear in those AI answers. Trades businesses with complete Google Business Profiles, consistent directory listings, and fresh website content are better positioned to benefit.
Should I be worried if my business disappeared from Google AI answers after August 14?
Model swaps can cause temporary volatility. If your business was appearing in AI answers before the swap and stopped after, check your Google Business Profile completeness, run a few test queries, and monitor over the next two weeks. If it does not recover, the model may be applying a stricter filter that your profile does not meet yet.
How is Gemini different from ChatGPT for local business discovery?
Gemini pulls local business data directly from Google Maps and has a 100% profile accuracy rate, according to SOCi's 2026 Local Visibility Index. ChatGPT pulls from Bing and has 68% profile accuracy. Gemini recommends 11% of local business locations, while ChatGPT recommends 1.2%.
What is the fastest way to improve my AI visibility after a model swap?
Start with your Google Business Profile. Fill in every field. Then check your directory listings for name and address consistency. Then add LocalBusiness schema to your website. These three steps address the inputs Gemini 3.7 Flash evaluates when deciding which businesses to recommend.