Yext Research library

Every paper and how they connect

21 papers on AI search and local visibility. Read each one, follow a topic across them, and see where each sits on the map.

AI citations

Brief · Jul 2026

With location on, AI cites the sources that know places

Arjun Sangwan

Across 155.5M citations and 1,623 brands, 80% of location-grounded AI citations pointed to listings or brand websites.

Four in five citations are influenceable 80% pointed to brand websites or listings.
MapQuest out-cited Wikipedia 5.8 to 1 About 6.0M MapQuest citations against 1.0M for Wikipedia.
Anthropic leans on reviews and social Its reviews-and-social share is 7.5 times OpenAI's.
The dip from 88.2% reflects the sample mix Anthropic answers were scanned 15 times more than the prior quarter.
Citations keep growing for a fixed panel Total citations rose 2.77x across three quarters for 770 brands.

Read the paper

Brief · Jul 2026

AI rewards specificity

Arjun Sangwan

Broad queries pull AI citations to local pages. Queries that name a service or product pull citations to dedicated intent pages.

Broad category queries favor local pages 64% of primary-category citations landed on local pages.
Named intents shift citations to intent pages 70% of intent queries cited intent pages more often than local pages.
Specific intent queries hit intent pages most Three example queries cited intent pages 84.1%, 86.2% and 89.1% of the time.
The pattern holds across models Each model cited the intent page more often than the local page for one example query.

89.1% Intent-page citation share for the ID photo service query
3.8x Lift in intent-page citations for one intent query over its broad query
70% Intent queries that cited intent pages more often than local pages
64% Primary-category citations that landed on local pages
41.0% Intent-page share for the brake parts query, the lowest example

Read the paper

Brief · Apr 2026

AI citation behavior across models: evidence from 17.2 million citations

Yext Research

Four major AI models cite different source types at different rates. Brand visibility depends on which model a customer asks.

Patterns vary more within sectors than across them Restaurants and specialty cafes show different citation profiles.
Claude leans on user-generated content Limited Control citations run 2 to 4x higher than other models across seven sectors.
SearchGPT favors official hotel websites Hotel sites receive 38.1% of its citations.
Gemini favors first-party websites Full Control share ranges from 22.4% in Hospitality to 54.0% in Organizations.
Healthcare shows the narrowest model spread Full Control ranges from 34.60% to 40.45%.

17.2M Distinct AI citations collected in Q4 2025 across four models
2-4x Claude's Limited Control reliance compared with other models
38.1% SearchGPT Full Control share in Hospitality
24.35% Claude's Limited Control share in Food & Beverage. Gemini's is 2.57%.
4.31x Citation occurrences per URL for websites. Listings sit at 2.46x.
25.62% No Control rate for non-profits

Read the paper

Does every AI read you the same? (1 min 44 sec)

Brief · 2026

When AI shops for you: the retail battle for citations

Adam Abernathy

265,000+ AI citations from 2,284 domains show retail citations concentrate in brand-controlled sources.

Brand-controlled sources lead retail citations About 87% come from websites, verified directories and structured listings.
Directory citations concentrate in a few sources The top ten third-party directories hold 52% of directory citations.
Gemini rarely cites Google Google appears in under 2% of Gemini's citations.
AI favors reference-style data Models lean toward structured sources.

265,000+ AI citations across OpenAI, Gemini and Perplexity
2,284 Unique domains behind the citations
87% Retail citations from brand-controlled sources
52% Directory citations held by the top ten directories
5% Citation overlap per model

Read the paper

Brief · Feb 2026

AI citation report: Q4 2025 analysis

Adam Abernathy, Yext Research

17.2M AI citations from Q4 2025 across Claude, Gemini, Perplexity and SearchGPT, broken down by sector and control category.

Owned and listing sources lead Full and Some Control sources make up 90.6% of citations.
Reviews and social are a small share Limited Control is 5.48%. News and forums are 3.92%.
Retail leads sector share 23.93%, ahead of Healthcare at 20.10%.
Claude leans on Limited Control sources It leads that share in all seven sectors.

46.89% Citations from websites under full control
43.71% Citations from listings under some control
5.48% Citations from reviews and social
3.92% Citations from news, forums and other sources
38.08% SearchGPT's full-control share of Hospitality citations

Read the paper

Does every AI read you the same? (1 min 44 sec)

Paper · Oct 2025

The 88% advantage for AI in financial services

Christian Ward, Adam Abernathy, Alan Ai

A 2.3M-citation analysis of financial services finds brand websites supply 47% of citations and directories add 41%.

Most citations are directly manageable Websites and controllable listings represent about 88%.
Financial brands hold high first-party control The authors tie this to regulation favoring authoritative sources.
Local pages add citation surface area Branch pages and advisor profiles contribute.
Generic sources shrink once location applies Reddit and Wikipedia appear at near-negligible rates.

2,280,650 Citations across Gemini, OpenAI and Perplexity
47% Citations from first-party websites
41% Citations from controllable third-party listings
88% Combined share that is directly manageable
53.93% OpenAI's share from third-party directories
425,000+ Citations from MapQuest and Google

Read the paper

Paper · Oct 2025

AI citations, user locations, and query context

Christian Ward, Anthony Rinaldi, Adam Abernathy, Alan Ai

6.8 million AI citations across Gemini, OpenAI and Perplexity show why location and query context should anchor a visibility plan.

Most citations come from controllable sources Brands can influence or manage the sources behind about 86%.
Websites lead objective local queries First-party sites supply over 40% for all three models on unbranded factual questions.
Subjective queries shift OpenAI to directories Directories reach 46.3% for branded subjective questions.
Location-level views reveal hidden variation A national 47% first-party rate could mask 70% rural and 20% urban.
Forums appear far less for local queries About 2% of citations.

6.8M AI citations, July 1 to August 31, 2025
86% Citations brands can influence or manage
52.15% Gemini citations from first-party websites
48.73% OpenAI citations from controllable listings
20,820 Unique domains cited

Read the paper

AI recommendations

Paper · Sep 2026

What AI knows can change which businesses it recommends

Arjun Sangwan

Complete, accurate location facts help AI match a business to a customer's needs. The facts AI mentions often differ from the fields locations fill in.

Mention rates do not track filled-in fields Filled-in shares and AI mention rates fit weakly.
Pool leads hospitality answers Pool appears in 83.41% of hospitality answers. Only 3.37% of hotels mark it as available.
Accessible entrances rarely appear in finance answers 75.73% of financial locations mark one. It appears in 0.69% of answers.
Memory can carry needs into later answers Examples from OpenAI, Anthropic and Google show earlier preferences may shape later recommendations.
Broad prompts capture one kind of question The study used broad prompts without memory.

83.41% Pool mention rate in hospitality answers
68.37% Breakfast mention rate in food and drink answers
35.95% Walk-in mention rate in healthcare answers. 0.66% of providers mark walk-ins.
0.69% Accessible-entrance mention rate in financial services answers
1,761,754 Answers analyzed across 384,076 locations and six answer systems
97.98% Hotels marking free Wi-Fi, the least mentioned hospitality fact at 55.44%

Read the paper

What can AI say about you? (1 min 42 sec)

Brief · Sep 2026

Four AIs, one question, four different answers

Anthony Rinaldi

Four leading AI engines chose the same top local business in 4% of shared queries. Profile depth and reviews go with agreement.

The same top pick is rare Unanimous in 4% of cases. All four engines named different top businesses in 48%.
Bigger candidate pools widen the gap Each engine named about 7 businesses. The four lists combined averaged 23.
Rural markets agree more often Three-or-four engine agreement reached 23.7% in rural markets and 12.5% in the densest markets.
Being named starts with legibility Location pages, cover photos and claimed profiles show the strongest lift.
Ratings help and do not decide rank The highest-rated business in a list ranked first 40% of the time.

4% All four AI engines selected the same first-ranked local business in 4% of shared query-market cases.
91% The winning business’s own website was cited in 91% of unanimous top-pick cases.
40% The highest-rated business in an AI returned list ranked first only 40% of the time.
2.9× The report associates location pages with a 2.9× relative lift in being named by AI.
48% Queries where all four engines chose four different top businesses
491,972 Shared local query-market cases in the four-engine comparison

Read the paper

Why don't AIs agree on who's best? (1 min 54 sec)What does AI check before it books? (1 min 47 sec)Does AI read your reviews? (1 min 45 sec)

Local search and rank

Brief · Sep 2026

What changed in Google's mobile organic results in May?

Mark Kabana

Mobile organic results became 48.7% more varied across nearby locations after mid-May 2026. The mobile 3 pack and Maps barely moved.

The break starts in mid-May The first clear weekly rise came the week of May 18, overlapping Google's May core update.
Searches that name a place rose most Queries naming a city or area rose 76.3% in the matched comparison.
The 3 pack and Maps stayed flat Mobile 3 pack changed 2.4% and Maps 2.8%.
The change held through June and July Mobile organic stayed above its earlier range.
One rank describes a market less well A single observed rank now represents the wider market less reliably.

48.7% Mobile organic results varied 48.7% more across nearby locations after mid-May 2026.
97.8% 97.8% of matched reports showed higher geographic variation in mobile organic results.
76.3% Geographic variation rose 76.3% for mobile organic searches that explicitly named a location.
15.6% Mobile organic geographic variation was 15.6% lower in August 1–9 than July 20–31.
8.16M Location-specific organic SERP snapshots, January 1 to August 9, 2026
13.8% Fall in center domain overlap

Read the paper

Why does my rank change a block away? (1 min 57 sec)

Paper · Feb 2026

The data synchronization effect

Aditi Mittal, Anthony Rinaldi, Christian J. Ward, Matthew Hupman, Arjun Sangwan, Adam Abernathy, Deeksha Reddy

Across 21.6 million local search results, businesses that actively manage their digital presence rank 2.71 positions higher within one mile.

Managed businesses rank higher within one mile +2.71 positions across 1,826,571 results.
The advantage grows with competition +6.20 positions in markets with 100+ businesses.
The advantage reverses at longer distances -1.22 at 3 to 5 miles and -0.57 at 5+ miles.
The effect holds across most industries Positive in 19 of 20 industries with 2,500+ managed observations.
Smaller brands gain the most +4.23 for small brands and +3.41 for enterprise brands.

21.6M Local search results across 7 sectors and 30 industries
+2.71 Average rank lift within one mile
+6.20 Rank lift in markets with 100+ businesses
19/20 Industries with positive results
96.5% Managed profile completeness. 82.6% for unmanaged.
342 Average reviews for managed businesses. 248 for unmanaged.

Read the paper

Who wins the search around the corner? (2 min 14 sec)

Brief · Aug 2025

Best practices will only take you so far

Michael Iannelli, Sneha Kuchipudi, Adam Abernathy

8.7 million Google results show review activity drives Local Pack visibility. Signal strength varies widely by industry and region.

Review engagement is the most consistent driver Review volume and prompt owner responses lead across industries and regions.
Hospitality favors quality over quantity Photo count correlated negatively with rankings.
Food and dining values recent ratings More than overall volume or completeness.
Weekend reply lag can cut visibility South and West were penalized most for slow responses.

8.7M Google search results analyzed
200+ Data points per business location
2,500 Most populous U.S. ZIP codes included
6 Industry verticals
4 Census regions compared

Read the paper

What earns a spot in Google's top three? (1 min 22 sec)

Paper · Aug 2025

Optimizing keyword selection for local SEO

Ariana Martino, Adam Abernathy

A data-driven method to pick 3 to 7 distinct local keywords per location, with competitor analysis that goes beyond page one.

Keywords should map to distinct topics Cover different intents.
Deeper results reveal hidden competitors Chicken and Asian cuisine brands outranked a burger chain on hunger-intent queries.
Near-me terms lead search volume
City and ZIP modifiers trail Far less common than near-me variations.

3-7 Recommended distinct keywords per location
4.5M Unique keywords in the dataset
40 Local competitors returned per Scout search
4 Industries tested

Read the paper

Brief · Jul 2025

Impact of SERP ranking on engagement

Arjun Sangwan, Anthony Rinaldi, Adam Abernathy

Local Pack and lower-ranked listing engagement across 51,000+ businesses. Regulated industries show flatter click-through curves.

The Local Pack captures most impressions The top three drew 58.6%.
Impressions fall sharply past the top three Down 52% at ranks 4 to 10 and over 70% beyond the top 10.
Healthcare and finance show flatter curves Users browse deeper and still act.
Lower ranks convert more views into actions Ranks 21 to 40 show a 66.5% impression-to-action ratio. Top results show 9.8%.

58.6% Impressions captured by Local Pack ranks 1 to 3
2.7x Median impressions for Local Pack businesses against first-page peers
52% Drop in impressions at ranks 4 to 10
66.5% Impression-to-action ratio for ranks 21 to 40
51,000+ Businesses evaluated

Read the paper

What earns a spot in Google's top three? (1 min 22 sec)

Profiles and presence

Paper · Oct 2026

In healthcare search, the review bar is set next door

Arjun Sangwan, Danny Gavin

Across 27 healthcare categories, map visibility tracks a provider's reviews against nearby competitors. The bar sits far lower in behavioral and developmental care.

The local bar varies widely by category A typical competitor has about 225 reviews in dental and vision care and about 4 in behavioral and developmental care.
More reviews than nearby rivals lifts top-3 odds Profiles with 4x the typical competitor's reviews reach the top 3 about 2.2 times as often as distance predicts.
A perfect 5.0 rating trails 4.5 to 4.9 Profiles rated 5.0 reach the top 3 at 0.8x expected. Profiles rated 4.5 to 4.9 reach it at 1.5x.
Fresh reviews and owner replies help everywhere A review within 90 days gives 1.5x. No review in a year gives 0.7x. Replying to 90%+ gives 1.5x.
A few reviews matter most in quiet care In behavioral care, 15 to 49 reviews gives 1.7x expected. The same count in dental care gives 0.5x.

27 of 27 Categories where more reviews than nearby competitors goes with more top-3 spots
2.2x Top-3 rate for profiles with 4x the typical competitor's reviews, against distance expectation
35% Behavioral and developmental care profiles with no Google reviews
692,855 Healthcare Google Business Profiles scanned
328,258 Scout map searches run May to September 2026, ads excluded
0.2x Top-3 rate for profiles with no reviews, relative to distance expectation

Read the paper

Brief · Aug 2026

Active digital presence goes with more customer actions

Yext Research

Actively managed locations appeared on more third-party platforms. Locations with broader presence and more frequent updates received more customer actions.

Managed locations show up on more platforms Found on 89.61% of checked platforms against 37.47% for comparison locations.
Wider presence goes with more actions Locations on 50+ live platforms had more total actions the next month than those on 1 to 20.
Frequent real updates go with more actions In every size group, more update days went with more actions the following month.
The gap held across industries Managed locations had higher platform presence in all 10 industries.

Read the paper

How does AI decide who to trust? (2 min 51 sec)How many platforms can find your shop? (1 min 57 sec)

Brief · Jul 2026

Good profiles are graded on a curve

Anthony Rinaldi

Across 500 brands and 600,000 U.S. locations, a profile's standing against nearby competitors predicted its customer actions. That standing moved monthly.

Nearby competitors set the benchmark A profile is judged against the businesses customers see beside it.
Better standing draws more actions Top-tenth locations drew roughly 185 actions per 100 for an average location. Bottom-tenth drew about 60.
The pattern holds across large brands It appeared in 476 of 483 brands with at least 200 locations.
Each signal has its own curve Reviews reward every gain. Photos punish falling far behind. Attributes plateau at local norms.
Higher standing goes with AI mentions Top win-rate groups were mentioned more often in AI answers. This is correlation.

600k+ U.S. locations across 500 brands and 13 industries
+1.37 SD Gap in customer actions between the best and worst profile tenths
98.6% Large brands showing the same standing-to-actions pattern
85% Extra actions for best-standing locations against an average location
53.9% AI mention rate for the top win-rate group. 42.3% for the lowest.
5 pt Yearly action gap between improving and declining weak locations

Read the paper

Who is your profile graded against? (1 min 40 sec)

Methods and data

Methods · Feb 2026

Elo rank study: how we got to the numbers

Yext Research

A walkthrough of how Yext applied Elo ranking to 21.6 million local search results, with every filtering step shown.

Managed listings rank higher nearby 2.71 positions higher within 1 mile.
The advantage fades with distance 1.26 at 1 to 3 miles. Negative at 3 to 5 and 5+ miles.
Competitive markets show the largest gap +6.20 in ultra-competitive markets.
Correlation only The study cannot prove management causes higher rank.

21.6M Local search results across 30 industries
+2.71 Positions managed businesses rank higher within 1 mile
-1.22 Managed rank difference at 3 to 5 miles
1,826,571 Within-1-mile results used for industry analysis
99.8% Managed businesses with claimed profiles. 77.9% for unmanaged.

Read the paper

Who wins the search around the corner? (2 min 14 sec)

Methods · Dec 2025

Converting the Scout Index to the H3 grid system

Mark Esposito, Ariana Martino, Aditi Mittal, Alan Ai, Adam Abernathy

The Scout Index moves from ZIP codes to the H3 hexagonal grid, at resolution 6 with 3,000 population-weighted sample locations.

H3 replaces ZIP codes ZIP areas vary in shape and are U.S.-only. H3 cells are near-uniform worldwide.
Resolution 6 is the working setting The coarsest level where rankings diverge from rooftop-level searches.
Square-root weighting balances countries Smaller countries receive more sample cells.
The sample skews to dense urban cells Lower-population areas may be added later.

3,000 Scan locations across the U.S., E.U., U.K. and Australia
~3.2 km Edge length of a resolution-6 hexagon
16 Resolution levels in the H3 grid
65M+ AI citations accumulated since July 2025
60,000+ Locations scanned since July 2025

Read the paper

Methods · Aug 2025

Building a monthly digital brand visibility observatory

Adam Abernathy, Lilla Szabo, Christian Ward

Background and methods for the Scout Index, a monthly observatory collecting over 2 billion data points.

Local ZIP-level scanning Keyword searches run in high-population U.S. ZIP codes.
Competitor mapping per keyword and location The top 40 competitors, plus their web and review profiles.
AI citation sources barely overlap AI models and traditional search return largely different domains.

2 billion Data points collected monthly
9.4M Business locations discovered
40 Top results retrieved per keyword in each ZIP code
4.5% Domain overlap between any two leading AI citation sources
1.3% Domain overlap between Gemini and ChatGPT

Read the paper

Journal · Jul 2025

Yext Research Journal, July 2025

Adam Abernathy

An introduction to Yext Research: a monthly brand visibility observatory, why generic SEO advice falls short, and local keyword selection.

A monthly observatory tracks local search shifts Time-indexed monitoring shows patterns static reports miss.
Generic advice has limited reach Google treats healthcare, restaurants and financial institutions differently.
Keyword count needs balance Too many terms dilute relevance. Too few reduce discoverability.

Every paper and how they connect

AI citations

AI recommendations

Local search and rank

Profiles and presence

Methods and data

Topics

Reviews

Brand websites

Listings and directories

Differences between AI models

How much brands control

Location context

Location and intent pages

Nearby competitors

Local rank

Complete, current profiles

Customer actions

Scout data

H3 grid

Managed listings

Healthcare

Financial services

Retail

Food and dining

Hospitality

Get the latest from Yext Research