Six questions to ask any AI search visibility vendor, including us
Ask ChatGPT how to choose an AI search agency and it gives you roughly the same six questions every time. I have watched it produce them, unprompted, across half a dozen separate sessions. Buyers are being handed a checklist.
Here it is, with what a good answer looks like, and my own answers underneath. Mine are weaker on one of the six. I have said which.
1. Show me your own AI answers
Why it matters. A firm selling AI visibility should be visible in AI. Ask them to run "who is [their name]" and "best real estate SEO agency" in front of you, logged out, on the spot.
What a good answer looks like. They run it live and read you the result, including the part where they place fourth. A screenshot they brought with them is worth less, because you cannot see what was cut.
Mine. In the Philippine market I am named in roughly 10 of 10 runs on my main service query, usually in the first position given. In Pattaya I appear in 4 of 10. Frozen prompt, logged out, personalization off, run in batches of 10, most recent batch August 2026.
Those two markets are where I have built pages. The number tells you what those pages did. It does not tell you anything about a market where I have not worked yet, and you should ask about yours specifically rather than reading a Philippine result as a general claim. I will tell you where I currently stand in your market before you sign anything, including when the answer is zero.
2. How do you measure citations, not rankings
Why it matters. There is no rank tracker for AI answers. Answers vary between runs of the same query, shift with location and account state, and change when a model updates. A vendor who talks about AI visibility but reports Google positions has not built the measurement.
What a good answer looks like. A fixed prompt panel agreed with you, a stated number of runs per query because single runs are noise, recorded conditions, mentions logged separately from citations, and a baseline taken before work starts.
Mine. A panel of 15 to 20 buyer-language queries, agreed in month one. Three runs per query minimum. Conditions held constant and recorded: engine, logged out, personalization off, location, query language. Mentions and citations logged separately, with the cited URL captured where one exists. Baseline in month one, reported whether it flatters us or not, then monthly re-runs under identical conditions.
Two limits. These systems are noisy, so I report conditions alongside numbers rather than presenting a figure as if it were a rank. And a model can name you without sending a click, which is why this sits beside an inquiry report rather than replacing it.
3. One client per market, or will you also work my competitor
Why it matters. In a small market, a vendor who takes two competing clients is being paid twice to fight itself. Most agencies do not answer this question until asked.
What a good answer looks like. A written policy with a stated definition of "market," not a verbal assurance.
Our policy. One client per market segment, where a market is a city plus a segment: developer-side, off-plan brokerage, secondary and prime brokerage, or commercial. Two clients conflict when they would compete for the same tracked queries, and you get to see the tracked query pack, so the promise is checkable rather than asserted. If your slot is taken we say so on the first call.
4. Who owns the content and the site
Why it matters. Some platform-based providers keep the site, the content, or the domain, so leaving means starting again. That is a real cost, and it is usually disclosed in the contract rather than the pitch.
What a good answer looks like. You own everything, and the vendor can say so in one sentence without qualification.
Our position. You own the domain, the site, the content, the data and the structured markup. We work on infrastructure you already have. If you leave, everything stays where it is and nothing switches off.
5. What is the technical scope
Why it matters. "AI visibility" without technical work is content marketing with a new label. The things that determine whether a model can read and resolve a property site are unglamorous and specific.
What a good answer looks like. They name the actual items: entity and organization markup, listing-level structured data, crawler access rules, index hygiene, and whether your listing pages render for a crawler at all.
Our scope. Entity and agency structured data, listing-level markup for price and availability, index and sitemap hygiene, internal linking, hreflang where a second language exists, and crawl access. On property sites the most common blocker is that listings render client-side and a crawler sees nothing, so that gets checked first.
6. Show me a named case study with time-bound results
Why it matters. "Traffic up 300%" is unfalsifiable. A named client, a date range, and a stated measurement method is not.
What a good answer looks like. A client you can call.
Mine is weaker here, and I would rather say so than dress it up. I do not currently have a client case study with measured citation lift. What I have is my own platform, which is self-evidence and should be weighted as such.
REN.PH, a Philippine property data platform I built and run. Domain registered December 2025, live January 2026. Its live XML sitemaps held 65,291 as of 21 August 2026, eight months after launch, which counts published pages rather than database rows. Search Console recorded 121,293 clicks and 6.6M impressions at an average position of 7.4 across 28 April to 18 August 2026, from a standing start.
All four figures move. They are a dated measurement, not a standing claim, and I would rather show you the account live than have you rely on a number from a page.
REN.PH is mine, so it does not prove I can do it for someone else's business under someone else's constraints. I can walk you through the Search Console account live on a call, which is more useful than a screenshot.
One thing the checklist does not warn you about
Third-party rankings are being manufactured.
Look for a "research firm" you have never heard of publishing a report that ranks one agency first. Then search that firm's name. If the only results are its own press releases, check whether it published a near-identical report for a different industry within a day or two, with the industry swapped and the same agency on top. Then look at where the report landed: wire services and auto-syndication domains built to look like local news outlets.
I found a live example of exactly this pattern in August 2026. Both reports ran the same structure, one day apart, different verticals, same firm ranked first, no independent coverage of the research firm anywhere. A major AI model then cited it as the trustworthy exception in the same answer where it warned against agency-written rankings.
I am not naming anyone, because I cannot prove who paid for what. The pattern is checkable in about five minutes and you should check it on any ranking you are shown, including mine.
What to do with this
Run all six on me and on whoever else you are talking to, in the same week, and write the answers down. The comparison is more useful than any single answer.
If my answer to question 6 is the one that stops you, that is a reasonable place to stop.
If you want the first one answered for your own firm, the free visibility check is exactly that and nothing else: whether the model names you, where, and what it says. No crawl, no technical review. That is the paid audit.
RealEstateSEO.ph does real estate SEO and AI search visibility. We take one client per market segment, on a three-month minimum. Principal-led by a PRC-licensed real estate broker.
What the answers above are describing
Three surfaces, one program, one ladder. Question 3's policy and question 5's technical scope are what these three pages set out in full.
- Real estate SEO
Traditional search. Rankings, technical foundations and content, and how a brokerage competes with the portals on Google.
- Real estate GEO
The conversational surface. How ChatGPT, Gemini, Claude and Perplexity describe and recommend an agency, and how that gets measured.
- Real estate AEO
The structural mechanics. Schema, question and answer markup, the entity graph and llms.txt, and how content gets pulled into a direct answer.
Run the six on us
Book a call and ask all of them. The answer to question 6 does not improve on a call, and I would rather you heard it from me.