Roughly 46% of Google searches carry local intent. A Google staffer said that at a conference in 2018, and marketers have repeated it ever since, usually without the date attached. Meanwhile, the audits that check whether local ad dollars reach actual nearby humans keep landing on a harsher number.
Contents
The audits nobody quotes in the pitch deck
Location Sciences reviewed 500 million impressions across the UK and US in 2019. Between 30% and 80% of location ad spend was wasted, the firm found. The data underneath it was inaccurate, poor quality, or fraudulent. Only 14% of tracked impressions carried genuine GPS data. Furthermore, 36% of the top GPS-enabled apps in that sample showed signs of location fraud.
Other vendors reached similar conclusions. Foursquare estimated that close to 80% of the location data in its bid stream was inaccurate, and it explained the incentive plainly. Passing location earns publishers a tenfold to twentyfold lift on ad yield. Placed found that only 1% of exchange-derived locations were precise enough to confirm a store visit. Average accuracy spanned more than four New York City blocks. Thinknear put it in simpler terms, since over half of location-targeted mobile ads missed by more than half a mile.
Read those numbers with one eye closed
Every figure above comes from a company that sells location verification. That conflict matters. Location Sciences and Mobilewalla published methodology, which puts them closest to primary evidence, while the rest circulated as press quotes. Still, no independent audit has contradicted the direction of the finding, largely because nobody has run one.
Apple shrank the panel
Apple's App Tracking Transparency prompt landed in April 2021 and cut the trackable pool immediately. Researchers studying billions of impressions across 19 countries found that US Apple traffic fell from 73% trackable to 18%, a drop of 55 points. Opt-in rates kept sliding afterward. Singular recorded 18.9% in the fourth quarter of 2023, then 15.9%, then 13.85% by the middle of 2024. Consequently, the panels feeding foot-traffic attribution now extrapolate from a thinner and less representative base than the dashboards suggest.
Regulators took another slice
The FTC spent 2024 dismantling the sell side of this market. X-Mode and Outlogic accepted the first ban on selling sensitive location data. InMarket, Gravy Analytics, and Mobilewalla followed, and Mobilewalla's order went further by barring the collection of bidstream data for anything beyond the auction itself. Kochava settled in 2026 after roughly four years of litigation. Because sensitive categories now cover clinics, places of worship, schools, and shelters, whole classes of geofence are simply off the table.
Even the market size is a guess
Sizing this category produces its own comedy. Stratistics MRC put hyperlocal advertising platforms at $3.2 billion for 2025. Fortune Business Insights valued hyperlocal services at $4.03 trillion for the same year. Those two numbers differ by a factor of more than a thousand. One counts ad platforms, while the other counts delivery gross merchandise value. Location-based advertising estimates cluster somewhat tighter, between $23 billion and $143 billion depending on the firm. Even so, tighter is doing heavy work there. Anyone quoting a hyperlocal market size without its definition attached is quoting nothing.
AI search is rewriting the top of the funnel
Local discovery started shifting at the same time. AI Overviews now appear on more than half of local queries, though the split by intent runs sharp. Informational local questions trigger them 92% of the time and hybrid questions 97%, while purely commercial queries sit near 8%. SOCi tracked consumers using ChatGPT to find local businesses at 6% in January 2025 and 45% a year later.
That growth has not translated into visibility for most operators. Only 1.2% of locations get recommended by ChatGPT and 11% by Gemini. Worse, just 68% of business contact details on those assistants match what sits in Google Business Profile. Practitioners at Whitespark argue that local search stays insulated, because assistants still serve stale hours and pricing. Both readings cannot be right, and the resolution matters for anyone budgeting past this quarter.
What survives the scrutiny
Ownable signals hold up better than bought impressions. Google Business Profile carries roughly 32% of local pack ranking weight and reviews another 16%, and both respond to work rather than spend. Review thresholds moved fast as well. BrightLocal found that 31% of consumers will only use a business rated 4.5 stars or higher, up from 17% a year earlier.
Local Services Ads also produce auditable outcomes. SearchLight Digital analysed $6.72 million in spend across 888 contractors and 126,650 leads. That work produced a blended cost per valid lead of $53. Electrical came in at $39 and roofing near $79. Furthermore, the book rate hit 43.9%, which produced a closed return on ad spend of 7.84 against an average ticket of $1,826.
Delivery economics punish density mistakes
Operators building hyperlocal delivery hit a different wall. Last-mile costs absorb 40% to 50% of logistics spend in Indian quick commerce, against average baskets of 450 to 600 rupees. Unlike traditional ecommerce, where batching improves the cost curve at scale, speed-constrained delivery inflects upward past a density threshold. Each extra order eventually demands another dark store or another idle rider. Blinkit crossed 50% market share by September 2025 and turned contribution-margin positive, while Dunzo shut down in January 2025. Density, rather than ambition, separated them.
Test instead of trusting the dashboard
Geo-experiments remain the cheapest correction available. Meta's GeoLift package builds synthetic control groups from untreated markets and runs power analysis before any money moves, which flags underpowered tests early. Google's Meridian handles regional marketing mix modelling and can accept experiment-derived returns as a prior. Neither tool needs device-level location data, so neither inherits the bidstream problem.
The practical rule follows from that. Judge a local channel on incrementality you measured, not on the visits a platform modelled for you. Google labels its store-visit conversions as estimates for good reason, and it attributes them at click time rather than visit time.
Where this leaves local operators
None of this argues against hyperlocal work. Proximity still decides purchases, and 72% of local searchers visit a business within five miles. The argument is narrower. Spend on the signals you own, and verify paid channels with geo-tests before you scale them. Then fix your listing data across assistants, and treat every footfall report as a hypothesis rather than a receipt.
By Callum Gracie, Otto Media.