Short answer
Fix access before content. Let the retrieval crawlers in, get indexed in Bing and Google, make your name, address, credentials and services agree everywhere, then claim and fill the directories that matter in your category. Writing comes sixth, because nothing you write gets read until the first five are true.
Key takeaways
- The order matters more than the effort. Five of these six are configuration and data, not writing.
- Bot protection at the CDN blocks AI crawlers more often than robots.txt does, and nobody tells you.
- One inconsistent phone number across four listings is enough to make an assistant describe a competitor instead.
- Published articles are worth doing, and they are last, not first.
Most practices start at the end: they commission articles. Then nothing happens for six months, because the site was never crawlable by the bots that build the answers, and the listings still say the old suite number. Here is the order that actually works, ranked by how quickly each step changes what an assistant says.
1. Let the retrieval crawlers in
Check robots.txt, then check the layer above it. Bot protection on a CDN blocks unfamiliar crawlers by default, and the ones you need are unfamiliar: OAI-SearchBot, ChatGPT-User, PerplexityBot, Perplexity-User, Claude-SearchBot, Claude-User, plus Googlebot and Bingbot. This is a settings change, and it can move an answer within days.
Training crawlers are a separate choice, and a reasonable practice can decline them. GPTBot, ClaudeBot, Google-Extended and Applebot-Extended feed model training rather than live answers, so blocking them costs you nothing in today’s citations. Do not confuse the two lists: the names differ by a few characters and the consequences do not.
2. Get indexed, and confirm it rather than assume it
Submit the sitemap in both Bing Webmaster Tools and Google Search Console, then search a distinctive sentence from your own homepage in both engines. If the page does not come back, nothing downstream matters. Indexation is a yes-or-no gate, and most practices have never checked the Bing half of it.
3. Make your entity data agree, everywhere
Put your practice name, address, phone, hours, credentials and service list side by side across your website, Google Business Profile, and every directory that carries you. Fix every disagreement, including the ones that look trivial. Suite numbers, "Dr" versus "DDS", and an old phone number are exactly what cross-checking catches.
This is the step that protects patients as well as rankings. In rater8’s April 2026 survey of 992 US adults, 66% of the patients who had used AI to research a provider had met incorrect information, and 60% trusted what they read without checking it. Wrong hours in a directory become a patient standing outside a locked door.
4. Claim the profiles that carry weight in your category
Assistants lean on third-party sources they can verify. For a clinical practice that means Google Business Profile first, then Healthgrades, Zocdoc, your state licence board, and the association directories for your specialty. Same name, same credentials, same description on each. Reviews come from real patients only, asked in person, never written for them.
| Step | Who does it | Effort | When it shows up |
|---|---|---|---|
| Crawler access | Whoever runs the site or CDN | One afternoon | Days to weeks |
| Indexation in Bing and Google | Whoever runs the site | One afternoon | Days to weeks |
| Entity data agreement | Practice manager | A week of admin | Two to six weeks |
| Directory profiles | Practice manager | A week, then monthly upkeep | Weeks to months |
| Schema that matches the page | Developer | A day | Weeks |
| Articles answering real questions | Whoever writes | Ongoing | Months |
5. Add schema that matches what is visibly on the page
Organization or LocalBusiness for the practice, Person for each clinician with links to their licence and association listings, and FAQPage only on the pages where those questions are actually visible. Schema that describes content a visitor cannot see is a liability, not a shortcut, and site-wide FAQ markup is the common way practices break this.
6. Then write, and write the way the evidence says
Now articles pay. The peer-reviewed study on this tested roughly 10,000 queries and found that adding real statistics, direct quotations and cited sources raised visibility in generative answers by roughly 30 to 40%, while keyword stuffing changed nothing. Write the numbers you actually have, attribute them, and date them.
One practical test before you publish anything: can a stranger lift a single paragraph from the page, quote it, and be correct? If the paragraph needs the three paragraphs above it to make sense, an assistant will skip it for a source that does not.
What this list will not do
It will not make a two-month-old domain outrank a hospital system. Vendor research across 2025 and 2026 keeps finding that how established a domain already is predicts citation better than any single on-page change, and that 40 to 60% of cited sources churn month to month. Run the six steps, then judge over three to six months against a fixed list of questions.
Questions people ask
How do I know if my CDN is blocking AI crawlers?
Ask whoever manages your hosting to check the bot-protection rules for OAI-SearchBot, PerplexityBot and Claude-SearchBot by name, and to look at the request logs for those agents. robots.txt can be perfectly permissive while the firewall above it returns 403 to every one of them.
Should I block GPTBot and ClaudeBot?
That is a fair choice to make either way. Those are training crawlers, so blocking them keeps your content out of future model training without removing you from today’s answers. Keep the retrieval crawlers allowed regardless, because those are the ones that decide citations.
How many directories does a practice actually need?
Fewer than most vendors sell. Google Business Profile, your state licence board, one or two specialty association directories, and the two or three booking sites your patients already use. Consistency across five profiles beats scattered entries across forty.
Is FAQ schema still worth adding?
Yes for extraction, no as a promise. Google has shown FAQ rich results only for well-known government and health sites since 2023, so expect clearer parsing rather than a visible search feature. Mark up only questions a visitor can read on that same page.
Where do articles fit if they are last?
They are the part that compounds. The first five steps decide whether you can be quoted at all; articles decide what you get quoted saying. Start them once access, indexation and entity data are done, then publish steadily rather than in one batch.
Sources
- rater8, 2026 Patient Choice Report (April 2026, n=992 US adults, vendor survey): 66% of AI users met incorrect provider information; 60% did not verify it.
- Aggarwal et al., "GEO: Generative Engine Optimization", KDD 2024 (peer reviewed): Statistics, quotations and cited sources raised citation visibility roughly 30 to 40%.
- OpenAI, crawler documentation (accessed 20 September 2026): The split between GPTBot for training and OAI-SearchBot for search results.
Published . Last reviewed .
Want this in your business?
Book a call and we will scope exactly which part of your operation an agent should take first.
Book a call with the team