Citations are the second-largest signal feeding the prominence pillar after reviews, and they're the one most teams treat as a one-time setup task instead of an ongoing operation. The first round of citation building takes a few weeks. The drift that erodes them takes years and silently caps map-pack ranking the whole time. This chapter covers both — the build, and the cadence that keeps the build clean.

“A business with 200 inconsistent citations ranks worse than one with 25 consistent ones. Volume isn't the signal. Consistency across the right sources is the signal. Every citation a search engine reads is a vote on whether your business is one entity or two.”

What counts as a citation

A citation is any public mention of the business's NAP — name, address, phone — on a third-party site. The link is optional; the structured NAP is the signal. Citations come in three shapes:

• Structured citations. Listings on directories with explicit NAP fields (Yelp, Apple Maps, Bing Places, industry directories like Avvo for lawyers or HealthGrades for doctors). The dominant form and the easiest for search engines to parse.
• Unstructured citations. NAP appearing in body text on local press, blog posts, partner sites, sponsorship pages. Lower volume signal per citation but higher trust because it's editorial.
• Aggregator citations. NAP submitted to data aggregators (Data Axle, Localeze, Foursquare/Factual) that propagate to hundreds of downstream directories. One submission, many derived listings — but it cuts both ways: a wrong submission propagates wrongly to all of them.

The 12 directories that actually matter

The local-SEO industry sells lists of 200+ citation sites. The reality is that 12–15 sources do the heavy lifting in Google's local index, and the long tail rarely moves the needle on its own. The core list, in rough order of weight:

• Google Business Profile — the source of truth. Already covered.
• Apple Maps / Apple Business Connect — second-largest map index by usage.
• Bing Places — Microsoft's local index. Powers Bing, Bing Maps, and Copilot's local answers.
• Yelp — feeds Apple Maps and Siri locally; high-trust user reviews.
• Facebook — high-trust because the brand identity is owner-managed.
• BBB (Better Business Bureau) — US trust signal; weak elsewhere but US weight is large.
• Foursquare / Factual — data aggregator; propagates to hundreds of downstream directories.
• Data Axle (formerly Infogroup) — second major aggregator.
• Localeze / Neustar — third major aggregator; required for full long-tail coverage.
• Yellow Pages / YP.com — declining but still indexed.
• Industry-specific top-3. Avvo for lawyers, HealthGrades for doctors, Houzz for home services, OpenTable for restaurants. The category-specific weight is large.
• Local press / Chamber of Commerce. Local newspaper directories and chamber listings carry disproportionate trust because they're locally-rooted.

The exact-match rule

The format the Google Business Profile uses is the canonical NAP. Every citation should match that format byte-for-byte. The places teams break this rule unintentionally:

• Address line 2. "Suite 200" vs "Ste 200" vs "#200" vs "Unit 200" — pick one and use it everywhere.
• Phone format. "(617) 555-0143" vs "617-555-0143" vs "+1 617 555 0143". Different directories enforce different formats; pick the format the GBP uses and adjust where directories allow it.
• Business name punctuation. "Joe's Bakery" vs "Joes Bakery" vs "Joe's Bakery Inc." — the GBP-listed legal/operating name wins.
• Street abbreviations. "Street" vs "St." vs "St" — match GBP exactly.
• City spelling. "Saint Paul" vs "St. Paul" — ditto.

The audit playbook

Three steps to a clean citation graph:

• Inventory. Search the business name in incognito with no location bias. The first 5 pages of results plus a Google Maps search produce the realistic list of where the business is currently cited. Add the 12 core directories above. Cross-reference any aggregator-fed long-tail directories.
• Compare against canonical. Open each listing and check the NAP against the documented canonical. Flag any mismatch — name, address line 1, address line 2, city, state, ZIP, phone, hours, website URL, primary category. Even a different category counts as a mismatch.
• Correct in dependency order. Start with the aggregators (Foursquare, Data Axle, Localeze) — corrections at the aggregator level propagate downstream over 4–8 weeks. Then correct the top-12 directories directly. Then walk the long tail; many of those will self-correct after the aggregator update propagates.

Building net-new citations

For new businesses or businesses entering a new region, the build order:

• Week 1: Google Business Profile, Apple Business Connect, Bing Places, Facebook. The four owner-managed canonical sources.
• Week 2: Three data aggregators (Foursquare, Data Axle, Localeze). Submission once, downstream propagation kicks off.
• Week 3: Yelp, BBB, Yellow Pages, plus the top-3 industry directories for the category.
• Week 4–8: Local press / chamber / city directory mentions. These are slower because they often require outreach (a press release, a chamber membership, a sponsorship). High-trust per citation.
• Ongoing: One new locally-relevant citation per quarter — partnerships, sponsorships, event listings, "best of" features. The slow drip that signals continued local presence.

The drift problem

Citations decay. The directories don't change your record, but the world around the business does:

• Phone migration. A VoIP switch, a new area code, a tracking number swap. Update GBP first, then the citation graph in the same week — never let the citation NAP and the GBP NAP diverge.
• Address change. Even a unit-number change is a major event. The full citation cleanup is a 6–8 week project; do not rush it. The map-pack hit during transition can last 60+ days even when handled well.
• Hours updates. Some directories propagate updates automatically; some don't. Verify the top 12 quarterly.
• Owner-name changes. Rare but devastating. A rebrand or a legal-name change requires submitting a name-change request to every directory plus a fresh aggregator update.

Citation tools — when they help and when they don't

The major citation services (Yext, BrightLocal, Whitespark, Moz Local) all do real work. The framing that decides whether to use them:

• Yext. Subscription-based, real-time updates across a large network. Strength: speed and breadth. Weakness: when you cancel, many of the listings revert to whatever they were before, because Yext is the publisher of record. Treat it as rented citations, not owned.
• BrightLocal / Whitespark. One-time submission services plus monitoring. Citations are owned (you submitted them) so they don't revert when the service ends. Slower than Yext but cleaner ownership model.
• Moz Local. Hybrid — submits to the major aggregators on your behalf and monitors the propagation. Reasonable middle ground.
• DIY. The 12 core directories take 6–10 hours of focused work. For a single-location business, DIY for the core + one service for long-tail propagation is the leanest setup.

How citations feed AI search

AI engines (ChatGPT, Gemini, Claude, Perplexity, AI Overviews) build their local-business knowledge from the same citation graph Google uses for the map pack, plus Wikipedia and Wikidata. When a user asks "what's the best pediatric dentist near me," the engine cross-references the citation graph to identify candidate businesses, then summarizes from the most-cited records. NAP inconsistency causes the engine to either pick the wrong record (citing an old phone number, an outdated suite, a closed location) or to disambiguate the brand into multiple candidates and lose confidence — which usually means the engine doesn't cite the brand at all.

The implication is that citation work, which used to be exclusively about the map pack, is now also AI-citation hygiene. The next chapter, local keyword research, moves from prominence into the relevance pillar — the queries the citations and profile need to be optimized against.