We Checked 3,000 Businesses Against Their Own Google Listing. 1 in 4 Disagreed.

Why nobody has this number

Citation consistency is standard local SEO advice, and the reasoning behind it is sound. Google has to decide which physical business a listing, a website and a directory entry all refer to, and agreement between them is evidence. Disagreement isn't.

What's missing is the base rate. Measuring it needs both halves at once: the verified Google Business Profile, and the live site. Rank trackers hold the first. Crawlers hold the second. We happen to hold both, for 23,732 businesses, because that's what our product does.

So this is a base rate, not a ranking study. We're not claiming to have measured what inconsistency costs you in the map pack. We measured how often it is there.

How we measured it

We sampled 3,000 businesses that carry a Google phone, a Google address and a website, and that we had successfully crawled before. For each one we fetched the homepage in a single pass and looked for the Google phone number three ways: in tel: links, in JSON-LD structured data, and in the visible page text.

Everything is compared on digits alone, last ten, so (862) 318-3997, 862.318.3997 and +1 862 318 3997 are the same number. Formatting never decides a verdict.

• Match — the Google number is on the page.
• Conflict — the page shows one or more numbers, none of them the Google one.
• Absent — the page shows no phone number at all.

319 sites did not answer and are excluded from every rate below, because a site that blocks a crawler isn't evidence in either direction. Most of those were a plain 403. A further 25 had no usable Google number to compare against and are also excluded.

Finding one: one in four disagrees with itself

The two failure modes aren't the same problem. A business with no phone on its website has an omission. A business publishing a second, different number has a contradiction, in the exact place a search engine looks to confirm the business is real.

The 15.2 percent is the group worth thinking about. These businesses are actively publishing two answers to the question "what's your phone number", and they're publishing both about themselves.

Finding two: the obvious objection, measured

Anyone who works in local marketing will raise the same thing immediately, and they're right to. Call-tracking software deliberately swaps the number shown on a website so that calls can be attributed to a campaign. A conflict on one of those sites is a working setup rather than neglect.

That objection is testable, so we tested it. We already detect CallRail as part of ordinary enrichment.

Call tracking is real, and it is 3.1 times over-represented among conflicts, exactly as the objection predicts. It also accounts for 34 of the 403. The other 91.6 percent have no call-tracking software on the page at all.

We detect CallRail but not CallTrackingMetrics, WhatConverts, Invoca, Marchex or DialogTech, so the genuinely explained share is higher than 8.4 percent. For it to account for the whole 15.2 percent, the vendors we do not detect would have to be roughly ten times more common among these businesses than the market leader, which isn't plausible.

Finding three: a third do not show their address

We looked for the ZIP code from the Google listing anywhere on the homepage. A ZIP is the cleanest single token in an address: five digits, rarely coincidental, and present in any real footer address block.

61.9 percent showed it. 38.1 percent did not. That's 994 businesses of 2,606 with a checkable address, whose website never states the postcode Google has on file for them.

Finding four: three quarters publish no structured data

25.9 percent of these businesses publish LocalBusiness structured data or one of its subtypes. The other 74 percent publish none.

This is the cleanest finding in the study because there's no ambiguity in it: no call-tracking equivalent, no formatting question, no judgement call. Either the markup is on the page or it isn't, and for three businesses in four it isn't.

These are companies with an address, opening hours, a phone number and a service area, none of it in a form a machine can read without guessing. The machine doing the reading is increasingly an AI assistant answering a question rather than a search engine ranking a page.

What to do with this

If you sell local SEO, this is a ten-second qualifying check. Copy the phone off the prospect's Google Business Profile. Search for it on their website. It fails 28.5 percent of the time. When it fails you are not guessing at a problem, you are holding two numbers the business published about itself.

Rule out call tracking before you call. Look for CallRail or a similar product in the page source first. On our numbers that is about 8 percent of conflicts, and walking into one of those makes you look like you didn't check.

If you run a local business, do the check on yourself. It costs a minute. There's roughly a one in four chance your own site disagrees with your own listing, and if it does you almost certainly don't know.

Our own interest here is straightforward: we sell data about businesses, and this is a signal we can compute because we hold the Google listing and read the website. We now store the Google phone and the website phone as two separate fields rather than reconciling them into one. A single phone column that quietly disagrees with what a prospect shows visitors looks like our mistake. Two labelled fields make it clear whose it really is.

Methodology notes for anyone citing this: 3,000 US businesses sampled from our Google Maps corpus on August 20, 2026, each carrying a Google phone, address and website, and each previously crawled successfully; homepages fetched in a single pass with a plain HTTP fetcher identifying itself as LyreLeadsResearch; phone numbers extracted from tel: hrefs, JSON-LD telephone fields and strictly-formatted visible text, compared on the last ten digits; 319 unreachable sites and 25 without a usable Google number excluded from all rates and reported separately; 2,656 judged. Call-tracking detection covers CallRail only, so the explained share of conflicts is a lower bound. Two caveats belong with any citation. This is a Google Maps corpus of the verticals we have searched, weighted heavily toward roofing and property management, so it is not a random sample of local business. And "absent" is an upper bound: a phone number drawn by JavaScript or set in an image reads as absent to a plain fetcher when a human would see it. Cite freely with a link.