Yelp, Thumbtack, and the New Data-Partnership Tier in AI Citations

Earlier this year we documented why ChatGPT local recommendations run through Foursquare's Places API rather than the directories most practitioners recommended. The 60-70% Foursquare signal was real -- and largely invisible because practitioners were tracking page citations, not API-level data queries. That held for most of 2025.

Three data partnerships signed between October 2025 and July 2026 added two more tier-1 structural signals to the stack. For home services businesses, Foursquare is no longer the only fix. For Claude local citations specifically, a direct path now exists that didn't exist 18 months ago.

What "data partnership" means versus organic citation

Most directory citation relationships work through organic training data and live web retrieval. Perplexity crawls Yelp pages. ChatGPT trained on Angi content. The AI finds a relevant page and extracts the business as a source.

Data partnerships work differently. The directory licenses its structured business database to the AI platform via API. The AI doesn't crawl the directory page -- it queries the database. A business absent from that database doesn't have a crawl-prioritization problem. It has an absence problem. Website quality and schema markup don't change what's in or out of the licensed dataset.

This distinction matters for fix plan logic. Organic retrieval gaps respond to directory page optimization, content quality, and crawl accessibility. Data-partnership gaps respond only to claiming and completing the profile on the licensed platform. These are different problems that need different fixes.

The Yelp + OpenAI partnership (July 2026)

Yelp licensed 330 million reviews and 8 million-plus business listings to OpenAI for direct use inside ChatGPT. Our August 2, 2026 methodology update (Scout sessions 95-96, confirmed from Yelp's primary press release) documented what this changes.

Before this deal, Yelp's relationship with ChatGPT was organic. Yelp content may have appeared in ChatGPT's training data, but there was no direct structured feed for local queries. After July 2026: Yelp is a Tier-1 ChatGPT signal, co-equal with Foursquare as a data-partnership source. A claimed Yelp profile now has its data inside ChatGPT's local response layer via the licensed database, not through organic retrieval.

Yelp also announced a "Request a Quote" integration for ChatGPT, still rolling out. When that launches, Yelp profile completeness won't just affect citation probability -- it will control whether a business appears as a bookable option directly inside ChatGPT. The incentive to claim and complete the profile is now double.

Thumbtack + ChatGPT (October 2025) and Thumbtack + Claude (April 2026)

Thumbtack integrated into ChatGPT via OpenAI's Apps SDK in October 2025 -- nine months before the Yelp deal. US users asking home services questions are served Thumbtack professionals from within ChatGPT via API. Our August 2, 2026 methodology update (Scout session 96) documented this as a vault gap: the Thumbtack + ChatGPT integration predated any practitioner recognition of Thumbtack as a structural ChatGPT signal.

Then in April 2026, Thumbtack's 300K-plus home services marketplace integrated directly into Claude -- Free, Pro, and Max tiers. Home services queries in Claude now surface Thumbtack professionals via API, not through web crawl.

The combined implication for home services businesses: a business not on Thumbtack is structurally absent from both ChatGPT and Claude for hiring queries. Not less likely to appear -- absent as a candidate in the data layer entirely. Foursquare presence, Yelp reviews, schema markup, and website content don't substitute for missing Thumbtack data.

This supersedes the earlier fix plan guidance that treated Thumbtack as a secondary directory. For home services audits, Thumbtack is now Tier-1 alongside Foursquare and Yelp, and it's the primary fix for low Claude scores.

The updated priority stack for home services

Our August 2, 2026 methodology update revised the ChatGPT and Claude fix sequences for home services audits:

**Tier 1 -- data partnerships (structural):** - Foursquare (December 2024 partnership) - Thumbtack (October 2025 for ChatGPT; April 2026 for Claude) - Yelp (July 2026 for ChatGPT; already Tier-1 for Perplexity and AI Mode)

**Tier 2 -- organic training-data directories:** - BBB, Angi

For non-home-services categories: Thumbtack doesn't apply (home services marketplace only). The Foursquare + Yelp data partnerships apply across categories. BBB and Angi remain relevant as organic training-data signals for services and non-services queries alike.

Why schema doesn't close this gap

Our August 6, 2026 methodology update (Scout session 99) resolved a research conflict on schema markup that had been in the vault since session 80. The short version: schema functions as entity-establishment infrastructure for businesses that have few or no AI citations, but does not increase citation frequency for businesses already receiving citations on some platforms.

For a home services business showing up consistently on Perplexity and Gemini but absent from ChatGPT and Claude local responses -- that's a data-partnership gap, not a schema gap. The fix sequence from the session 99 update: claim and complete the data-partnership directories first. Schema cleanup belongs in a secondary pass, after Foursquare, Thumbtack, and Yelp profiles are accurate and complete.

Schema matters for businesses starting from zero citations across all platforms. It's part of the foundational infrastructure that needs to exist before other signals can work reliably. But for the specific pattern of "strong on some platforms, invisible on ChatGPT and Claude," the Tier-1 fix is directory data, not markup.

What the Thumbtack profile needs

Claiming the profile is necessary but not sufficient. The Thumbtack API surfaces professionals based on profile completeness and query match. The fields that determine whether a business appears for relevant queries:

**Service categories** should use the most specific Thumbtack taxonomy option available. "HVAC repair" beats "Home Services." The API matches by category, and parent categories are too broad to surface correctly on specific service queries.

**Service area** must be defined explicitly. The integration matches professionals to location queries based on declared coverage, not inferred proximity.

**Pricing range** should be set, even as a broad estimate. Home services queries frequently include cost qualifiers, and an empty pricing field is a completeness gap.

**Reviews** are the floor. Our methodology (sessions 69, 77) documented 4.0-plus stars and 80-plus reviews as the threshold for reliable AI citation probability on Yelp. Thumbtack-specific thresholds haven't been confirmed separately, but a minimum of 5 reviews is the current practitioner floor for a profile to surface in AI responses.

What this means for the Yelp profile

Everything that made Yelp important for Perplexity still applies -- claimed profile, accurate categories, complete hours, current information. The Yelp + OpenAI partnership adds ChatGPT to the list of platforms where Yelp profile quality has structural impact.

The review floor (4.0-plus stars, 80-plus reviews from sessions 69 and 77) applies to ChatGPT via the data feed the same way it applies to Perplexity. A business in the Yelp database with 14 reviews is present, but may not surface in AI responses that filter by quality signals within the licensed dataset. Being claimed is the entry point; review volume and rating are what determine whether the profile actually surfaces.

If you want to see where your business stands -- which platforms are citing you, which aren't, and whether data-partnership gaps are driving the pattern -- a Signal Check at sourcepull.ca will show the current state across ChatGPT, Claude, Perplexity, and Gemini.