Skip to content

Evidence

What the research shows about getting recommended by AI

No independent study has yet measured a local business’s leads before and after this kind of work, and we won’t pretend otherwise. This page keeps four kinds of evidence apart (published research, what we’ve observed ourselves, labeled samples and verified client results) so you always know which is which.

Part 1 of 4 · Published research

What independent research shows

A controlled academic experiment, two large industry datasets, and one controlled test that shows what doesn’t work. Each links to the original so you can check every number.

Study 1 · Controlled experiment

Up to ~40%

more visibility in AI answers from adding quotations, statistics and cited sources

GEO: Generative Engine Optimization

Aggarwal, Murahari, Rajpurohit, Kalyan, Narasimhan & Deshpande, Proceedings of the 30th ACM SIGKDD Conference (KDD 2024). Princeton University, Georgia Tech, Allen Institute for AI, IIT Delhi. · 2024

The researchers rewrote the same web pages in nine different ways, ran a large benchmark of real search queries through generative engines (including Perplexity), and measured how much of each AI answer drew on each version of the page.

  • Adding quotations from credible sources+41%
  • Adding statistics+31%
  • Citing sources+28%
  • Keyword stuffing−8%

Limits: Figures are relative lifts in position-adjusted word count, computed from the paper’s main results table. The test used a research benchmark, not local businesses specifically, and results varied by topic.

What we do with it: Every page we write includes specific numbers, named local projects and attributed quotes, and we never keyword-stuff.

Read the original study

Study 2 · Large-scale correlation study

3×

stronger link to AI visibility for brand mentions across the web than for backlinks

An Analysis of AI Overview Brand Visibility Factors (75K Brands Studied)

Ahrefs (Louise Linehan & Xibeijia Guan) · May 2025

Ahrefs measured how strongly a range of search factors correlated with how often 75,000 brands were mentioned in Google’s AI Overviews.

  • Branded web mentions (correlation)0.664
  • Branded anchor text0.527
  • Branded search volume0.392
  • Number of backlinks0.218
  • Brands with zero AI Overview mentions26%

Limits: Correlation, not causation. The study covered established websites (Domain Rating above 40), not small local businesses specifically.

What we do with it: Earning mentions on local news, association and partner sites is a core part of our Growth and Enterprise work, not an add-on.

Read the original study

Study 3 · Large local dataset

7 in 10

local AI citations come from business websites and Google Business Profiles

What 200k local AI searches tell us about what businesses get recommended

BrightLocal (Mike Hawkes) · September 2026

BrightLocal ran about 200,000 non-branded local prompts for more than 1,300 business locations through ChatGPT, Google AI Mode and Google AI Overviews, and analyzed about 1.97 million citations.

  • Citations from business websites42%
  • Citations from Google Business Profiles28.6%
  • Citations from Yelp9.5%
  • Local searches where ChatGPT named a tracked business33%
  • Same, on Google Maps66%
  • Overlap between repeat ChatGPT runs of the same prompt23%

Limits: Run by a vendor using its own tracking tool. It describes where AI looks; it does not test before-and-after changes.

What we do with it: Your website and Google Business Profile get fixed first. Because answers change between runs, we track each question twice per platform every month and report a score, not a single screenshot.

Read the original study

Our honest stance on hidden site code

Behind-the-scenes code is hygiene, not a magic switch

Ahrefs compared 1,885 pages that added JSON-LD schema between August 2025 and March 2026 with 4,000 matched pages that didn’t. Citations moved +2.4% in Google AI Mode and +2.2% in ChatGPT, both indistinguishable from random noise, and −4.6% in Google AI Overviews. Every page studied was already being cited by AI, so the test can’t say whether schema helps pages that aren’t yet visible.

We still add this behind-the-scenes code, accurately, to every client site, because it is cheap, standard and helps search engines read the page. But we treat it as basic hygiene, not the reason you get recommended, and we will never sell it to you as a magic switch.

Ahrefs, “We Tracked 1,885 Pages Adding Schema. AI Citations Barely Moved.” (May 2026)

Also worth knowing

Why this matters now

Part 2 of 4 · Our own observations

What we’ve seen ourselves

Real public facts we checked on local business websites and profiles, and the same checks run on our own site. Small samples, not research, and businesses are never named. Missing from an AI answer says nothing about the quality of anyone’s work.

Anonymized public findings

  • Fort Myers Roofer (Anonymized)5-star real reviews, but homepage code read 2.0 stars to AI crawlers.
  • Cape Coral HVAC Shop (Anonymized)4.9 stars and 800+ reviews, zero website linked on Google Business Profile.

Our own site, same checklist

6 technical checks on www.wrightfordbackoffice.com, last verified Nov 20, 2026.

  • · Behind-the-scenes business details for AI
  • · Matching email, location and service area
  • · Clean HTTPS, no outdated HTTP links
  • · Plain-English service and FAQ pages
  • · Same prices and terms on every page and in our chat
  • · Search and answer tools can read the public pages

Nov 21, 2026 baseline (v1.1): named in 0 of 30 unbranded question–platform pairs on ChatGPT, Perplexity and Gemini (0 of 60 answers). Asked about us by name, Gemini described us but listed the wrong industries; ChatGPT and Perplexity couldn't verify us. Updated monthly.

Part 3 of 4 · Illustrative samples

What a snapshot looks like

These show the format of our work using an invented business. They are not evidence that the work succeeds, and we never present them as client results.

ILLUSTRATIVE SAMPLE — NOT A REAL CLIENT RESULT.

Sample AI Visibility Snapshot

The full report format: what was tested, how AI recognized the business (mentioned, described, cited, recommended), the sources behind the answers, and practical next steps.

Open the sample snapshot →

ILLUSTRATIVE SAMPLE — NOT A REAL CLIENT RESULT.

Example Walkthrough

The short version on our homepage, with demo numbers that use the same math as a real snapshot.

See the Example Walkthrough →

Part 4 of 4 · Verified client results

Results from real clients

Published only with the client’s written permission, always with the sample size and the like-for-like comparison behind each number.

None published yet

We’re early, and we won’t fill this space with case studies no one can check. Once a client has a full 90-day comparison and agrees in writing, their result appears here with its numbers and denominators. Until then, the most useful evidence is your own: the free snapshot shows what AI says about your business today.

Curious how AI describes your business today?

The free AI Visibility Snapshot tests 10 real buyer questions for your city and trade (for example “best roofer in Fort Myers for a tile roof replacement” or “emergency AC repair in Cape Coral”), each asked twice on ChatGPT, Perplexity and Gemini: 60 attempted AI checks (10 × 3 × 2). Any answer a platform doesn’t return is shown as unavailable, never counted as a miss. Plus a full check of your website and the listings that matter for your business, such as Google Business Profile, BBB and your industry’s directories. You keep the findings whether or not you hire us. Already decided? Start the month-to-month Core plan, and Jacob starts on your fixes this week.