dental bridgereturns nothing
the keyword, and no metrics field at all
One of fifteen dental terms the planner declined to answer. An empty row is not a zero, and Google is not short of data — its suggestion endpoint lists terms with ten searches a month.
tooth bridge12,100 searches
$13.38 average cost per click
The same appliance, a different word, and the planner answers in full. Neither term is ambiguous, and no reading of the refused list admits one and refuses the other.
Thirty-four dental terms asked of Google’s own keyword planner: eighteen came back with numbers, fifteen with the keyword and nothing else. Rename a refused term and you usually get silence back — twelve times out of fourteen. Twice it leaked cleanly, and this is one of them.
Root canal. Tooth extraction. Dentures. Dental bridge. Cavity filling. Emergency dentist. Wisdom teeth removal.
The interesting question is not which fifteen. It is whether the silence tells you anything about the procedure, or only about the words you typed. I spent a second round of measurements finding out, and the answer is uncomfortable in both directions.
The silence follows meaning, and I can prove it against myself
Rename a silent term and ask again. I sent fourteen paraphrases, one keyword per request: root canal therapy, endodontic treatment, endodontist, tooth removal, dental extraction, oral surgeon, denture, denture repair, emergency dental care, 24 hour dentist, walk in dentist, tooth filling, tooth bridge, tooth crown.
Two came back with numbers. Twelve stayed silent.
That matters, because across my original list Google priced 18 of 33 — a 45 per cent silence rate. Among synonyms of terms it had already refused, the silence rate is 86 per cent. A Fisher exact test puts that at p = 0.022. But those fourteen strings are not fourteen independent tries — they are seven families: three ways of saying root canal, three of saying extraction, and so on. Collapse them to one per family and the same test gives p = 0.095. So call it a strong lean rather than a proven effect: knowing that a string means the same as a refused term tells you it will probably be refused too.
So the flag is not blind to meaning. It clusters. Every string in the root-canal family — root canal, root canal treatment, root canal therapy, endodontic treatment — is silent. So is every string I tried with "emergency" in it, and so are the two walk-in phrasings I invented to test it, 24 hour dentist and walk in dentist.
But I cannot tell you what the cluster is
The obvious description is "restorative, surgical and urgent". It does not survive the table below, and neither did anything else I tried: procedure against practitioner, elective against needed, insured against cash-pay, word count, or the presence of the word "dental".
dental implants is surgical and restorative, and it is the most expensive click in the set at $26.00. gum disease is a pathology nobody elects, and it prices at $2.75. Meanwhile the root canal — the textbook restorative procedure — is refused under every name I tried: root canal, root canal treatment, root canal therapy, endodontic treatment.
And the cluster leaks. Twice, cleanly:
Those are the same appliance and the same complaint. Neither tooth bridge nor toothache is ambiguous. There is no reading of "restorative" or "urgent" that admits one and refuses the other.
Three more near-misses point the same way without being as clean. broken tooth is silent while chipped tooth prices at $16.35 — a chip and a fracture are not quite the same injury, so call that suggestive. root canal is silent while endodontist prices at $9.43 — but that is a practitioner rather than a procedure, and practitioner nouns are not a safe route either: six of the fourteen I tried are refused, including oral surgeon, implant dentist and emergency dentist. And in the non-dental control set, computer repair prices while laptop repair does not.
So: the silence is predicted by meaning, strongly, and then breaks on cases meaning cannot explain. That is the honest shape of it, and it is why the practical advice below is a hit rate rather than a rule.
What Google says about it
There is an explanation. It is two sentences, filed under troubleshooting rather than anywhere a planner would read first.
Sometimes. As a precaution. Not a category, not a list you can look up, not a rule you can plan around. That is a poor answer, and it is also the only one consistent with what I measured — a cluster with holes in it is exactly what "sometimes" produces.
It is not a rule about medicine
Google does treat health as a sensitive category elsewhere: its personalised advertising policy names invasive medical procedures, cosmetic surgery and injections outright. So the theory has real policy behind it. The control set kills it anyway:
Botox and a hair transplant are elective cosmetic work and they price; liposuction, which sits beside them on any list of elective cosmetic work, does not. Cataract surgery is not cosmetic at all. Neither are laptop repair and phone repair, which are refused while computer repair and car repair answer normally. All four refused controls stay refused when sent one keyword per request.
Four more things it is not, each with a file behind it
Not geography. The same fifteen sent to the United States, the United Kingdom and Canada return an identical silent list. The geo setting demonstrably works — dentist reports 1,000,000 searches in the US, 135,000 in the UK, 110,000 in Canada — and the silence does not move with it.
Not the network setting, and not a flaky response. Search-only and search-with-partners return the same fifteen. Three identical calls in a row return the same fifteen.
Not batch size. I was sending twenty keywords per request, and Google does merge near-identical keywords inside a request: I asked for thirty-four terms and got thirty-three rows, because dental implant was absorbed into dental implants as a close variant. If the silence were a side effect of that merging there would be no finding here, so I sent every refused term again, one keyword per request — the fifteen dental ones plus the four silent controls below. All nineteen still returned nothing.
Not a shortage of searches. This is the dull explanation and it deserved a real test. Google's planner has a second endpoint, the one that suggests keywords rather than pricing the ones you name. Seeded with dentist, dental, tooth, teeth, dental procedures and dental office, it returned 2,733 ideas, 2,732 of them carrying a search volume — the smallest at 10 searches a month.
My first instinct was to note that none of the refused terms appears in that list and call it proof. Checking it properly showed that is too strong: seven of the eighteen terms Google happily priced are missing from the list too, including invisalign at 368,000 searches a month. A suggestion list is ranked for relevance and truncated, so absence is not on its own a refusal.
The asymmetry is real, though. Fifteen of fifteen refused terms are absent against seven of eighteen priced ones, which a Fisher exact test puts at p = 0.00017. Absence leans heavily the same way as the refusal — it is just not proof by itself.
What does mean something is what the list *does* contain:
Seven of the fifteen refused terms have a priced phrase built on top of them — twenty-seven such phrases in all. Google will tell you that tooth extraction cost is worth 18,100 searches a month and then decline to say anything at all about tooth extraction.
So the data plainly exists. Whatever the refusal is, it is attached to the exact string and not to a gap in Google's records — and one extra word is sometimes all it takes to get around it.
What the eighteen priced terms actually cost
United States, English, Search only.
| Keyword | Searches / month | Top-of-page bid | Average CPC |
|---|---|---|---|
| orthodontist | 246,000 | $6.22 – $37.49 | $23.44 |
| dental implants | 165,000 | $6.55 – $29.14 | $26.00 |
| crowns | 301,000 | $1.33 – $22.77 | $1.85 |
| dentist near me | 1,830,000 | $4.86 – $22.08 | $13.48 |
| dentist | 1,000,000 | $5.17 – $21.58 | $13.78 |
| dental checkup | 33,100 | $4.93 – $20.02 | $12.20 |
| kids dentist | 60,500 | $3.79 – $18.95 | $12.96 |
| pediatric dentist | 201,000 | $3.69 – $18.21 | $11.15 |
| braces | 201,000 | $2.87 – $17.50 | $17.23 |
| periodontist | 110,000 | $3.71 – $17.37 | $10.18 |
| dental cleaning | 22,200 | $4.14 – $16.81 | $13.26 |
| cosmetic dentist | 110,000 | $3.40 – $16.58 | $14.50 |
| teeth cleaning | 14,800 | $3.34 – $15.53 | $10.88 |
| clear aligners | 18,100 | $3.42 – $14.95 | $13.21 |
| invisalign | 368,000 | $3.01 – $11.61 | $10.22 |
| veneers | 201,000 | $2.56 – $10.52 | $7.86 |
| teeth whitening | 165,000 | $1.86 – $7.33 | $5.42 |
| gum disease | 110,000 | $0.80 – $2.00 | $2.75 |
Read the two columns carefully, because they are not what most people assume. Google defines the top-of-page range as approximations of the 20th and 80th percentiles of what advertisers "have historically paid" for a keyword's top of page bid. So the high figure is a percentile, not a ceiling, and one row sits above it: gum disease averages $2.75 against a $2.00 eightieth percentile. That is one row in eighteen, and one in eight across my control set. I do not know why, and neither column claims to bound the other.
These are converted figures. Keyword Planner answers in the currency of the account you ask from, and the request has no currency field I could find — currencyCode, currency and currencyConstant all come back 400 · Cannot find field. The account is Ukrainian, so every dollar above is converted from hryvnia at the National Bank of Ukraine rate for 27 August 2026, 44.5717 to the dollar.
One row behaves unlike the other seventeen
crowns shows 301,000 searches a month, a top-of-page eightieth percentile of $22.77, and an average click of $1.85 — a twelfth of it. The other seventeen dental rows land between 0.59 and 1.37 on that same measure. crowns sits at 0.08.
Its monthly series peaks in October, at 450,000 against a 301,000 baseline. October is Halloween, and a crown is a costume before it is a restoration.
I want to be careful about how much that carries, because it is easy to oversell.
The seasonal factor is not special: veneers peaks by almost exactly the same multiple, 1.498 against crowns' 1.495 — and Google reports volumes on a fixed ladder that steps by about 1.22, so 1.50 is simply two rungs. What differs is the month, not the size. Nor is a low ratio proof of anything on its own: in my own control set plumber sits at 0.40, car repair at 0.45 and running shoes at 0.55, all below the dental floor, and nobody thinks a plumber's traffic belongs to someone else. And ambiguity alone does not produce it — braces is genuinely ambiguous, orthodontic and knee and suspenders, and sits at 0.98.
I have volume and price for this term and nothing else: no search-term data, no conversions. So crowns is a row worth being suspicious of, not a diagnostic you can apply.
What the pages ranking for this query tell you instead
I read the pages ranking for "google ads for dentists". Ten results: one serves a Cloudflare challenge to automated readers, one is a video whose page yields 257 characters of interface furniture, so eight were read. Six of the eight sell dental PPC management to practices; one sells a white-label version of it to other agencies; one is a general digital agency.
Three name a cost per click, and a fourth works a hypothetical at a $5 bid and a $3 click. Set against the planner's average CPC, which is the comparable number:
The agencies are not simply wrong. Implants sit squarely inside the Canadian range, and the broad dentist figures sit inside the $5–$15 one. Two rows miss on the same side: the Canadian general-dentistry row runs roughly two to six times below what the planner reports for the same cleanings and checkups, and the "$2 to $12" range tops out below both.
What none of them does is tell you where the number came from, or that the tool you would use to check it goes quiet on a great deal of dentistry. Not one of the eight mentions Keyword Planner. Not one mentions that some terms come back empty. Only one prices clicks by service category at all, and six of the eight never use the word "extraction".
This is the fourth one
I have now written four articles about Google Ads data, and three turned out to be about something Google does not hand over.
Auction insights exists inside Google Ads and cannot be automated: twelve fields were removed from the connector, and the API marks the metrics not publicly available. The search terms report withholds low-volume terms, then accounts for them in a second surface that reports seven metrics, none of which is cost. Conversion numbers keep moving for weeks after you send the report — that one is timing rather than withholding, but of thirty-five templates Google publishes itself, exactly one mentions it.
Now the planner returns no numbers for fifteen of the thirty-three dental rows it sent back, and the sentence explaining it sits under troubleshooting.
None of these is a scandal. Each has a reason, and privacy is a real one. The pattern worth naming is where the disclosure lives: never on the screen showing the number, always a page or two away, in wording that assumes you already suspected something.
What to do about it
Treat silence as unknown, not as zero
This is the one that costs money. An empty answer looks like a keyword nobody wants. It is not: Google will price tooth extraction cost at 18,100 searches a month while refusing tooth extraction. The demand is there, and it belongs on a dental practice scorecard whether or not the planner will admit it.
Rename and ask again, but budget for it failing
Two of my fourteen paraphrases worked. That is worth five minutes and not worth a plan. Ask singly, because Google merges near-identical keywords inside a request and a batch can hide the answer.
Try adding a word before you try a synonym
It has the better hit rate here: seven of the fifteen refused terms have a longer phrase built on them that Google prices without complaint. tooth extraction is refused and tooth extraction cost is not. Synonyms worked twice in fourteen; the extra word worked seven times in fifteen.
Anchor agency quotes on average CPC, not the top-of-page range
The range is a 20th-to-80th percentile spread of what was paid, wide enough to make almost any claim look reasonable, and none of the pages above says which figure it is quoting.
Be suspicious of a very low average against a very high percentile
crowns averages a twelfth of its eightieth percentile. That is not proof of anything by itself — three of my control terms sit low too — but it is worth ten minutes with the search terms report before you commit a budget.
For terms that stay silent under every name, ask the auction
A tightly matched ad group, a low bid cap, and read the real cost per click after a hundred clicks. Budget honestly: at what the planner reports for cleanings and checkups, $10.88 to $13.26, a hundred clicks is $1,088 to $1,326. That is a guess by analogy, because the planner will not price the terms you are testing — and the one agency that publishes an emergency figure puts it above general dentistry, not below.
What I could not establish
I do not know why any particular string is flagged. Google says "sometimes, as a precaution" and that is the whole of the public explanation. I can rule out geography, the network setting, instability, batching, the medical category, and a shortage of searches. I cannot rule out meaning — the paraphrase test says meaning matters — but I also cannot name the category, because it admits dental implants and refuses dental veneers.
I cannot tell you which rename will work. Two of fourteen did, and I see no rule in the data for predicting which.
Everything above is one advertiser account. I cannot rule out that a different account sees a different fifteen, and if you check yours and get another list, that is a more interesting finding than mine.
And this is the API, not the interface. I have not checked whether Google's own screens refuse the same strings, or refuse them in the same way.
What my own template does with this
Google Ads Vitals reads one campaign-level export — date, campaign, campaign type, impressions, clicks, cost, conversions, conversion value — and puts cost per lead, ROAS and revenue against targets you type in.
It has nothing to say about keyword planning, because keyword-level data is not in that export. If you want to know what a click costs before you spend, the planner and a small test budget are the tools, not my file. What comes out of your practice management system is a separate question, worked through in Open Dental and Dentrix reports.
What it does do is tell you what your clicks actually cost after the fact, blended across the account — the same kind of number the pages above are all estimating, except measured.
Questions people actually ask
For the terms Google will price, the average cost per click on 27 August 2026 ran from $1.85 to $26.00 — though that floor is crowns, a row the article gives reasons to distrust. The broad terms — dentist at $13.78, dentist near me at $13.48 — in the middle. Specialities do not cluster: orthodontist averaged $23.44, but periodontist $10.18 and pediatric dentist $11.15 both came in below the broad terms. These are Keyword Planner’s own averages for the United States, converted from the account currency, and they describe a market rather than your account: what you actually pay depends on the auction, your quality score and your competitors on the day.
Google’s only public explanation is on support page 9247190, under “Troubleshooting failed keyword attempts”: with sensitive topics it works to avoid showing ads where they may be inappropriate, and “because each search is different, keywords are sometimes flagged as a precaution.” Measured rather than read, the silence clusters by meaning — twelve of fourteen synonyms of already-silent terms were also refused, against a 45 per cent base rate — but it does not map onto any category you could name. Dental implants is surgical and prices at $26.00; dental veneers is purely cosmetic and returns nothing.
No, and the difference costs money. The API returns three distinct outcomes: a row with metrics, a row with no metrics field at all, and no row. A missing metrics field means Google declined to answer that string. It is not a shortage of data, and the clearest proof is that Google prices phrases built on the refused terms: tooth extraction returns nothing while tooth extraction cost reports 18,100 searches a month, and dentures returns nothing while dentures price reports 14,800. Seven of the fifteen refused terms have such a phrase. Reading an empty answer as a zero makes a keyword look worthless when it is simply undisclosed.
Rename it and ask again, one keyword per request — Google merges near-identical keywords inside a single request, so a batch can hide the answer. Budget for it failing: two of the fourteen paraphrases tested here came back with numbers. Adding a word does better — seven of the fifteen refused terms have a longer phrase Google prices, such as tooth extraction cost at 18,100 a month. If a term stays silent under every name, the only way to price it is a small live test: at what the planner reports for cleanings and checkups, a hundred clicks costs $1,088 to $1,326.
WRITTEN BY
Olha · clinic data analyst
I build the reporting our managers open every morning at a multi-branch medical clinic — and package it so other practices don't have to start from scratch.
Published on 27 August 2026. Three limits worth stating outside the body text. This is one advertiser account, and I cannot rule out that another sees a different fifteen — if you check yours and get another list, that is a more interesting finding than mine. This is the API and not the interface, and I have not looked at what Google’s own screen does with these strings. And the ranking pages are a snapshot of one index on one day: DuckDuckGo’s results are substantially Bing-derived, so this is a picture of a search market rather than the search market. An earlier version of this piece read the fifteen silent terms as a list of procedures Google will not price; the paraphrase test described in the sources is what showed that was wrong. I sell a template that reports what clicks cost after the fact, which is exactly why every figure above is one you can reproduce from the method listed.