You can do the first layer of GEO yourself: check whether the site can be read, record a small set of patient questions and fix clear website problems. Hiring help is a later decision about recurring measurement and coordinated implementation, not a requirement for getting started.
DIY GEO means owning the diagnosis, a small question set and the website tasks that follow. It does not mean treating one AI answer as a score or assuming that a technical fix causes a recommendation.
What can you do yourself this week, free?
You can establish a usable baseline this week without paying an agency: test whether the site can be read, ask a small set of patient-style questions, save the answers and turn clear website faults into tasks for the person who controls the site.
- Run the free checker. Use the AI visibility checker to get a first reading of accessible site issues.
- Write down five real questions. Use the treatment, location and concern that patients actually use; keep the wording and date with each result.
- Check the raw page, not just the browser view. Ask the web developer whether important copy is present in HTML and whether crawler access is allowed. Google's robots.txt guidance explains the permission-file layer.
- Check basic structured data. Confirm that the practice's visible identity and service information agree with its markup; Google's structured-data documentation is the primary reference.
- Create a short, owned task list. Give specific access, copy or data corrections to the person who can implement them; do not make changes merely to chase a single answer.
What does the 527-practice dataset show?
| Observed condition | Count and denominator | DIY use |
|---|---|---|
| Open to every AI crawler tested | 492/527 (93.4%) | Check access first; do not assume it from a normal browser visit. |
| No LocalBusiness or Dentist schema | 284/527 (53.9%) | Compare visible practice details with the markup before adding anything. |
| Both LocalBusiness and FAQPage | 18/527 (3.4%) | Use visible questions and answers; markup is not a substitute for clear content. |
Source: Citevio's open data library, practice-web-anatomy counts CSV, June 2026. Google also documents how JavaScript rendering affects search.
The access layer is worth checking yourself precisely because it is the one nobody looks at. Citevio’s own live access test requested 499 dental homepages with AI crawler identities on 22 July 2026. Every one of those sites had a robots.txt that welcomes those crawlers. 72 of them (14.4%) were refused by their own server anyway, and in 12.6% of the same sample Googlebot was served normally in the same run — meaning a classic SEO audit, which reads the permission file rather than the server’s answer, shows nothing wrong at all.
It is one day’s photograph: hosting and CDN rules change without notice, and a repeat on another day will not reproduce the file exactly. It measures access, not citation, and removing a refusal was never measured to produce a mention. Which layer caused the refusal — host, security product, CDN, plugin or a human decision — was not recorded. And it is a different instrument from our robots.txt study, which reads what a file declares across 6,497 domains rather than what a server does. The two rates are not interchangeable and cannot be averaged.
Source: Citevio live crawler access study, 22 July 2026. n = 499 sites whose robots.txt allows AI crawlers and whose homepage loaded for a plain browser request. Read it in full in robots.txt open but AI bots blocked.
Where does DIY stop working?
Where doing it yourself stops working is not a question of skill. It is three practical things:
- Once is not a measurement. A single check tells you what one engine said on one day, and these answers move between runs. A reading becomes evidence only when the same questions are asked on a schedule and the results are kept.
- A reading is not a decision. The result tells you where you stand; it does not tell you which page to change or what to publish next. That translation is the actual work, and it is the part that does not finish on its own.
- A badly run test is worse than no test. A question phrased differently each time, or a single run treated as a verdict, turns noise into a confident conclusion — and a confident wrong conclusion produces worse decisions than never having looked.
What we will not tell you is how many hours a week this takes. We have not measured it, and a number invented for a sales page would be exactly the kind of claim this comparison exists to argue against.
DIY stops being a small project when no one owns a stable question set, the same results are not recorded over time, or the fixes span content, technical access and business information with no one accountable for their order. Those are coordination limits, not proof that an owner or office manager lacks skill.
A one-time audit can still be the cheapest sensible start: document the baseline, fix clear access or information problems, then decide whether a recurring review is worth owning. The question set is fixed and agreed with the client at the start, and logged. It does not change mid-month, because a set that changes cannot be compared month to month.
How does DIY relate to SEO?
Keep ordinary SEO work running when it is serving its purpose. Dental SEO and GEO overlap around a usable, crawlable site but answer different measurement questions. Your current SEO agency may also be able to help; see what an existing SEO agency can already do.
Use the free checker for a first reading of your website and then keep the questions and the result together before changing anything.
Prefer to question the engines by hand? See how to check what ChatGPT says about your practice.
How we measured this
This is a decision guide, not a performance study. Tool prices were read from the vendors' published pages on 25 August 2026. The dental-site figures on this page are from Citevio's June 2026 dataset; they describe the scanned sites, not an outcome caused by any one change.
Common questions
Can an office manager run a DIY GEO check?
Yes, an office manager can keep a small question log and coordinate clearly scoped fixes. They should not turn one changing AI answer into a performance conclusion.
Is DIY riskier than doing nothing?
DIY is useful when the result is documented and acted on carefully. It becomes risky when inconsistent questions or one-off answers are used to justify expensive changes.
What is the cheapest way to start?
Use the free checker, keep a small fixed question set and resolve obvious access or information issues before buying software or services.
Should I hire an agency or do GEO in-house?
Start in-house if someone can own a small, stable set of patient questions, record repeated results, and assign clear website tasks. Consider an agency when the work becomes recurring measurement and coordinated implementation across technical access, content, and business information. This is a responsibility decision, not a claim that an in-house owner lacks skill.
Can I do GEO myself?
Yes. You can check whether the site is readable, log a small set of patient-style questions, save the answers, and turn clear faults into tasks for the person controlling the website. DIY does not mean treating one answer as a score or assuming a technical change caused an AI recommendation.
Is a one-time audit a better starting point than an agency retainer?
A one-time audit can be the cheaper sensible start when it documents a baseline and identifies clear access or information faults. Fix those first, then decide whether recurring review is worth owning. It does not replace ongoing measurement where the practice needs comparable results and coordinated decisions over time.
What can I do for AI visibility this week for free?
Run the free checker, write down five real patient-style questions, keep the wording and date with each result, and inspect the raw page rather than only the browser view. Then turn clear access, copy, or information faults into an owned task list. These steps establish a baseline; they do not promise an AI mention.
What part of GEO cannot be handed off to an agency?
The practice must still own the business questions worth asking and decide who may act on the findings. An agency can help design, repeat, and interpret a measurement process, but it cannot substitute for agreement on patient questions or responsibility for the website changes that follow.
Why does this comparison refuse to invent a weekly-hours estimate for DIY GEO?
This page does not give a weekly-hours estimate because Citevio has not measured one. The time depends on the question set, who controls the website, and whether fixes span content, access, and business information. A sales-page number without that measurement would present an unmeasured claim as a fact.
Which five patient-style questions should I log for a DIY baseline?
Choose five questions that use the treatment, location, and concern patients actually use for your practice. Keep the set small enough to repeat and keep every question's wording unchanged while you compare results. The page does not prescribe universal questions because the baseline should reflect the practice's own services and patients.
Why should I keep the wording and date with every DIY result?
Keeping the exact wording and date makes later readings comparable. If the question changes, a different answer may reflect the changed input rather than a changed result. The date also records which reading is being compared, so a practice does not mistake one answer on one day for a continuing pattern.
Should I inspect raw HTML instead of relying on the browser view?
Yes. Ask whether important copy is present in the raw HTML and whether crawler access is allowed, because a browser can assemble content after scripts run. The browser view alone does not establish what a crawler receives. Use the raw-page check as one input, then assign any clear issue to the person who controls the site.
What does 492 open sites out of 527 tell a DIY checker?
In Citevio's June 2026 practice-web-anatomy counts CSV, 492 of 527 US dental practice websites were open to every AI crawler tested. For a DIY checker, that makes access a specific item to test rather than an assumption from a normal browser visit. The count describes the scanned sites; it does not show that access creates an AI mention.
Does an open robots.txt file prove a server will allow an AI crawler?
No. An open robots.txt file states a permission-file rule, but the server can still refuse an AI crawler. Citevio's 22 July 2026 live access test found server refusals among sites whose robots.txt welcomed the tested crawlers. Check the actual crawler path and status response rather than treating the file as proof of live access.
When does one AI check become a measurement?
One check becomes a measurement when the same questions are asked on a schedule and the results are kept together. A single answer only records what one engine said on one day. Repetition with a stable question set makes it possible to distinguish a repeated observation from a snapshot.
Why is a changing question set not comparable over time?
The question set is fixed and agreed with the client at the start, and logged. It does not change mid-month, because a set that changes cannot be compared month to month.
How should I assign clear website faults after a DIY check?
Turn each clear access, copy, or data fault into a short task assigned to the person who can implement it. Record what was found and avoid changes made merely to chase a single AI answer. Ownership matters because a list without a responsible person does not resolve a website problem.
When does DIY become a coordination problem rather than a skill problem?
DIY becomes a coordination problem when nobody owns a stable question set, results are not recorded over time, or fixes span content, technical access, and business information without a clear order or owner. Those conditions describe responsibility gaps, not evidence that an office manager or practice owner lacks the skill to contribute.
Why is a badly run AI test worse than no test?
A badly run test turns noise into a confident conclusion. Changing the wording, treating one run as a verdict, or failing to save the result can lead to expensive changes based on a reading that cannot be compared. No test leaves uncertainty visible; a weak test can make an unsupported decision look justified.
Can I keep ordinary SEO running while I do a DIY GEO baseline?
Yes. Keep ordinary SEO work running when it is serving its purpose. SEO and GEO overlap around a usable, crawlable site, but the DIY baseline asks a different question: what the selected AI engines return for a stable set of patient-style prompts. Treat the baseline as an added measurement layer, not an instruction to stop useful SEO work.