Getting mentioned in ChatGPT comes down to two things you can control: whether the engine is allowed to read you at all, and whether your pages contain anything specific enough to be worth quoting. Everything else sold in this category is either unproven or actively counterproductive, and we will show you the measurement that proves it. What nobody can offer you is a guarantee, and by the end you will understand exactly why that is not us being modest.
A note on sourcing. Everything below is tied to platform documentation from the companies running these systems, or to peer-reviewed and preprint research. Search volumes come from our own keyword research, 1,341 healthcare keywords analyzed, July 2026.
Why this matters right now
A patient sits on the edge of the bath at eleven at night, phone in hand, asking an assistant where to go about a tooth that has been aching for three days. One practice gets named. Maybe three do. Everyone else in that city may as well not exist for the length of that conversation.
That is a new kind of shelf space, and unlike the map pack there is no visible ranking to climb and no ads to buy your way into. It gets decided somewhere you cannot see, by a system whose owners publish very little about how it chooses.
Which is exactly the condition under which a market fills with confident people selling certainty. So before any tactics, here is the uncomfortable part, stated plainly.
The promise nobody can make
A 2026 academic survey reviewing three years of research into optimizing for generative engines reached a conclusion that should be printed on the wall of every agency selling this: no reviewed technique shows a stable, longitudinal, cross-platform causal effect on organic discoverability.
Read it carefully, because the wording is precise. Researchers can demonstrate that once your content has been retrieved, certain properties change whether it gets cited. What nobody has demonstrated is a reliable way to get you discovered in the first place.
That is the thing being sold. It is the thing we are telling you cannot be sold.
We include ourselves in that. We can measure where you stand, fix what is verifiably broken, and do the content work the evidence supports. We cannot promise ChatGPT will name you, and anyone who does is either not reading the research or counting on you not to.
What does not work
Buying mentions across the web. The pitch sounds reasonable: engines trust third-party sources, so we will place you on a network of sites. Google addresses it directly, noting that seeking inauthentic mentions across the web is not as helpful as it might seem. There is a real observation underneath the pitch, which is that independent coverage does show up disproportionately in AI answers. The bulk-purchased version is not that, and the research describes the earned-media finding as observational, not causal.
Publishing forty thin pages. Google’s guidance asks for material with a unique point of view rather than recycled information, and warns specifically against scaled content abuse, meaning pages made primarily to move rankings. Volume was never the lever.
Keyword stuffing. It transfers from old SEO as a penalty rather than a shortcut. The survey found it reduces the position-adjusted visibility metric, with multiple benchmarks confirming null or negative effects.
Buying a recipe. One benchmark tested 54 combinations of method and domain and found only three significantly positive, none in question answering. Gains also erode as more people adopt the same trick, which is the nature of anything sold at scale.
Optimizing your pages for AI, in isolation. This is the one that should stop you. In end-to-end testing, optimizing page bodies alone reduced average top-20 presence by around 9%, top-10 presence after reranking by 16%, and final citation by 6%. Tuning a page to be quotable, without regard for whether it still ranks, damaged the retrieval that would have delivered it to the model. People who “did AEO” and watched traffic fall were not unlucky.
What the evidence supports
| Lever | Strength | What it means |
|---|---|---|
| Answering the question asked | Strongest, most reproducible | Models favor alignment with the question over credibility signals |
| Being retrievable at all | Foundational | Ordinary search visibility sits upstream of everything |
| Quotable specifics | Consistent gains | Statistics, definitions, comparisons, prices, dates |
| Not blocking the crawler | Binary | Checkable in ten minutes, free to fix |
Answer the question, in their words. The survey found query relevance the most reproducible lever tested, and noted that models favor alignment with the question over credibility signals. A plain page that answers what someone asked beats a more impressive page that talks around it. Most practice websites are written to reassure the owner rather than answer the patient, which is the single biggest reason they never get quoted.
Be findable first. Position in the retrieved context mattered more than most rewrites. To be shown as a supporting link in Google’s AI features, a page must be indexed and eligible to appear in Google Search with a snippet, with no additional technical requirements. The unglamorous work is the work.
Give the engine something to lift. Statistics, definitions, comparisons, prices, dates, references. The survey reports quotation addition producing around a 41% relative gain in the foundational study. Publish your actual prices and your actual hours and the specific conditions you treat, and you have handed an engine a sentence it can use with your name attached.
Do not lock the door. ChatGPT’s search appearance is governed by a crawler called OAI-SearchBot, which is separate from GPTBot, the one that governs training data. Sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers, though they can still appear as navigational links. We find this broken more often than anything else, and nobody ever chose it. A security plugin tightened its rules, or a bot blocklist grew over the years.
What to do in the next two weeks
- Check the crawler. Read your robots.txt, confirm OAI-SearchBot is not blocked, and confirm your CDN is not blocking it either. About 30 minutes.
- Build your prompt list. Twenty questions your patients ask, written the way they would type them. Freeze the list. About 1 hour.
- Run it and record everything. Ask three or four assistants, note who gets named, and note which sources the answers cite. That source list is the most useful thing you will produce all month. About 2 hours.
- Answer your top five questions properly. First two sentences of the page, the answer itself, not the preamble. About 1 hour per page.
- Add the specifics. Prices, timeframes, conditions treated, what happens at the first visit. About 2 hours.
- Rerun the list monthly. One reading is weather. The pattern is what tells you anything.
Roughly a day and a half of work, and the only cost is your time.
So what’s really working?
Not a tactic. A position, and a habit.
You cannot make an engine mention you. You can make yourself the easiest thing in your city to name: readable by the crawler, aligned with the question, full of specifics worth quoting, and consistent enough across the web that a machine knows what you are. That is the whole honest list.
Then there is the part almost nobody does, which is the habit. Ask the engines the same questions every month and write down what comes back. It costs nothing, it belongs to you, and it converts this from a field of confident claims into something you can observe.
That matters more than it sounds, because getting named is not the finish line. Research measuring citation quality found only around 51.5% of sentences fully supported by their sources and 74.5% of citations correct. Engines name businesses and then describe them wrongly. And what one engine says predicts little about another, with roughly a quarter domain overlap found between two major surfaces. There is no single AI search to win.
So the operators who get further here are not the ones who bought the best tactic. They are the ones who can tell you what four engines said about them last month, and the month before, and what changed. Everyone else is guessing confidently.
We would rather hand you the log than the promise.