One category kept 64% of its results across cities. The other kept 13%.

How small this sample is, before any number. Two queries, four locations, one provider, one day, desktop only — eight captures for the comparison. That is enough to demonstrate a contrast between two categories and not enough to publish a rate, a trend, or a claim about SEO in general. If you want the strong version of this finding, it needs more queries, more cities and repeat captures over time. We have not done that.

The two tables

Overlap is the Jaccard index on the top-10 domain sets, each city measured against the national capture. 100% would mean the same ten domains in that city as nationally; 0% would mean no domain in common.

seo services

Only two domains persisted across every location we sampled. Chicago returned agencies that appeared nowhere in the national result; Denver returned a different set again. A local pack was present in all four captures.

small business website design services

The same seven domains held in all three cities. A local pack appeared in one capture of four.

What we think this means, and what it does not

What the data shows. In this sample, two commercial service categories that read as equally national behaved very differently by geography. One is effectively a set of local competitions wearing a national phrase. The other is a genuinely national field.

What it does not show. It does not show that SEO service terms are always local, that web-design terms are always national, or that either pattern holds outside these four locations on other days. Two queries cannot support any of that.

Why it mattered to us. We were choosing which market to compete in. On volume alone the answer looked obvious — one of these terms carries roughly seventeen times the monthly searches of the other. After measuring, the high-volume term turned out to be dozens of local competitions we have no local presence to enter, and the smaller one turned out to be a stable national field. The smaller number was the better opportunity, and only the geography test showed that.

Method, in enough detail to repeat or attack

• Instrument: SerpApi, Google, device=desktop, gl=us, hl=en, explicit location per capture.
• Locations: United States (national), Dallas TX, Chicago IL, Denver CO.
• Date: all captures 28 August 2026. Single day, deliberately — comparing across days would confound geography with volatility.
• Depth: top 10 organic per capture, Google-redirect entries excluded.
• Overlap: Jaccard on domain sets, city vs national. Domains, not URLs — a domain ranking a different page is still the same competitor.
• Raw: every figure recomputed from the stored JSON responses rather than copied from an earlier summary.

A calibration note worth more than the headline. We also ran the same queries through a second, cheaper search API on the same day. Against SerpApi it agreed on 73% and 50% of top-10 domains for the two stable queries — and only 20% for seo services, the localized one, where it returned a different top three entirely. Cheap search APIs are least reliable exactly where localization matters most. If you are measuring localized SERPs with a tool that has no location parameter, you may be measuring your tool.

Known limits

• Two queries. Four locations. One day. One instrument. Desktop only.
• No mobile captures — mobile SERPs can differ materially and we did not test them.
• No repeat captures, so day-to-day volatility is unmeasured and could account for part of the movement.
• Provider-reported positions, not a browser in each city.
• Ad counts are what the provider returned at capture time and vary by auction.