Methodology · last read November 18, 2026
How Delve Stats is built
Every number on this site is a Google estimate, pulled the same way each month and dated on the page.
Sources
- Prices and searches: DataForSEO Labs keyword overview, which carries Google Ads keyword data. Location United States, language English. For each search it gives monthly searches, the low and high top-of-page bid, average cost per click and 12 months of search history.
- Who shows: live Google results reads through DataForSEO's SERP API, desktop, located in the metro's main city, for each metro's most searched term. We record the Local Services Ads block and the regular paid ads when they appear.
Which metros are listed
We started from each state's largest metro areas (two to eight per state) and checked every candidate search for each. A state's top metro is always listed, ranked by searches in a typical month: each pooled search counts once, at its median month, so one spike can't lift a metro. Its second and third metros are listed when Google reported numbers for at least three of their searches. The first eight metros (six in Texas, two in Arizona) stay listed. We first ranked on 12-month averages, then switched on November 18, 2026, when spikes turned out to drive Greenville, South Carolina's total: Sacramento, Cincinnati, Myrtle Beach and Springfield, Illinois came in, and Inland Empire, Greenville and Toledo left (their old addresses redirect to the state page).
Which searches are listed
For each metro we build candidate searches from 40 common injury phrasings (personal injury lawyer and attorney, car accident, car wreck, truck and 18 wheeler, motorcycle, work injury, slip and fall, wrongful death and more) times each city name. Some metros use more than one city (Dallas-Fort Worth, Tampa Bay, Raleigh-Durham), and city names that several states share carry their state (“springfield mo”, “columbia sc”). Each metro lists up to 15 searches, ranked by monthly searches.
Google reports one pooled number for close variants of the same search. “injury lawyer greenville sc” and “greenville sc injury lawyer” both show the same volume. Each row shows Google's number as reported, so the rows overlap and should not be added up.
Wherever we add searches up (a metro's monthly total, the state and national totals), each pool counts once. Two searches are one pool when Google reported the same volume and the same 12-month history, matched by calendar month, even if their bids differ slightly or their 12-month windows end in different months. Until November 19, 2026 we matched only identical volumes and bids, which counted some pools two or three times; totals published before then were too high (about 44% across all metros, read November 18, 2026).
Medians work the same way: every median (metro, state and national, for top-of-page bids and average cost per click) counts each pool once, using the pool's most recent 12-month reading. Until November 19, 2026, medians ran over every listed row, so a group Google returned several times got several votes. Correcting that moved 57 of 104 metro medians by 5% or more; the national median high bid went from $258 to $266.
This month we checked 5120 searches. 844 had reported numbers and are listed; 4169 were left out.
The honest limit
Google withholds numbers for some searches, including big head terms. “personal injury lawyer dallas” returns empty, for example. We list only searches with reported search volume and bid data, and never fill a gap with a guess. Each metro page shows which searches were withheld.
All figures are Google estimates. Top-of-page bids are Google's range for what advertisers have bid to show above the organic results; average CPC is Google's estimate of the typical cost per click. Neither is what any named firm paid, and no client account data from anyone is used.
Local Services Ads rotate between identical searches. We read each metro's top search up to three times, stop at the first read that shows a Local Services Ads block, and otherwise keep the read with the most paid ads. When no block appears, the page says “not measured” instead of guessing.
Flags you will see
- $1,000+: the source's top-of-page bids stop at exactly $1,000. On November 18, 2026, 46 of 1,108 keywords sat at that ceiling and none above it, so a bid there is shown as “$1,000+”, and a median that includes one is a floor.
- Thin data: a metro with fewer than 3 distinct searches Google reported (close variants and repeated copies of one keyword group count once), or under 1,000 searches a month across them. Until November 19, 2026 this counted every listed row, with a minimum of 5, so copies of one group could lift a metro out of thin. Thin metros are shown, hatched on the map, and left out of the national rankings and “What moved”.
- Spike: a search whose 12-month history has one month at 10 times or more its typical month (Google reported “personal injury attorney midland” at 110, then 33,100, then 170 searches). Its yearly average overstates normal demand. Metros where over a quarter of the searches come from spikes are left out of “What moved”.
- Not reported: Google gave no average cost per click for that search. We leave it blank instead of estimating.
The national pages
Seven pages roll the metro numbers up nationally. Each uses only the monthly read on this site; nothing is pulled separately, and each page carries its read date and a Dataset description for search engines and AI readers.
- Cost by state: medians across every listed search in a state's metros. A state is ranked only if at least one of its metros is not thin (fewer than 3 distinct searches, or under 1,000 a month). The CSV holds the same table.
- Most expensive searches: the top 25 by average cost per click, counting only searches with 100+ a month, because low-volume searches swing widely. Close variants Google pools together count once per metro.
- Cheapest markets: metros and states ranked from the lowest median top-of-page high bid. Thin metros and states are left out, so a few low-volume searches can't make a market look cheap.
- Most expensive markets: the same ranking from the top. It ranks whole metros and states by their median; the most expensive searches page ranks single searches.
- Truck vs car accident: median average cost per click for each case type. It uses average CPC rather than top-of-page bids, because many truck bids sit at Google's $1,000 ceiling. A state is compared only with at least 3 distinct truck and 3 car searches that have a reported CPC, and so is a metro. Anything with fewer on either side is listed separately as single readings, with no ratio. (The floor was 2 for states and 1 for metros until November 19, 2026.) The page also carries a second read, taken November 19, 2026 straight from Google Ads (DataForSEO's Google Ads search volume endpoint, United States, English) for truck and car searches in the 20 metros with the most injury searches. It is shown only as a 20-metro median on each side, because no single metro had enough truck searches to break out.
- Search demand: average monthly searches per metro with close variants counted once, and searches by calendar month across every tracked metro, over the months all of them report. Searches whose peak month is 10 or more times their typical month are left out, so a one-month burst can't rank a market.
- Car accident cost per click: median average cost per click for car accident searches, nationally and by state. A state is ranked with at least 3 such searches that have a reported cost per click, and each median shows how many it comes from.
Case types come from the words in each search, first match wins: truck accident, motorcycle accident, car accident, slip and fall, dog bite, wrongful death, workplace injury, and everything else is general injury or accident (truck is checked before car, so “truck accident lawyer” never counts as car). A case type needs 10 searches to be ranked.
The “what moved” email
Each refresh keeps the previous month's read, and the monthly email compares the two: which searches' top-of-page high bid rose or fell 5% or more, and who showed in Local Services Ads. Bids at the $1,000 ceiling can't show a real change, so they are left out of moves. When nothing moved 5%, the email still shows the biggest small moves, labeled as small. The first issue has no earlier month, so it shows the month's numbers instead.
The trend charts
Each search's 12-month history can end in a different month, so histories are lined up by calendar month before they are added, and only months every search covers are shown. Close variants share one pooled number, so each pool is counted once. Google rounds monthly volumes into coarse steps (9,900, 12,100, 14,800 and so on), which makes single months jumpy, so the bold line is a 3-month average, and the dots are the monthly totals as reported.
Refresh
The data refreshes automatically early on the first Monday of each month (about 6:00 AM Central). Each page carries its read date. The most recent read was November 18, 2026; Google last updated the keyword numbers between September 16, 2026 and November 18, 2026.