Apply one test to every question on your FAQ page: could ChatGPT answer this correctly with no internet access at all? If the answer is yes, that question will never earn you a citation, because nothing about it requires anyone to find your page.
The test, and why so many questions fail it
An AI assistant answering a question does one of two things. It answers from what it already knows, or it goes and looks something up. Only the second path produces a citation.
So the question decides everything. "What is a pre-approval?" is answered from training. "How does escrow work?" is answered from training. "What does a buyer's agent do?" is answered from training. A model has read those explanations thousands of times and has no reason to go hunting for an agent's website in a mid-sized city to confirm them.
Now change the shape:
How long does a condo sale usually take to close in this city, and where do the delays actually happen?
What has strata insurance done to condo affordability here over the last two years?
A buyer offered over asking and the appraisal came in below. What normally happens next in this market?
Which neighbourhoods here still have detached homes under a specific price, as of this month?
No model knows those from training. They are local, current, or situational, and answering them requires finding a source. That is the moment your page becomes eligible to be that source.
The uncomfortable part is that most agent FAQ pages are made entirely of questions that fail the test, because they were built from a template or copied from a national site. The page is not underperforming. It was never in the running.
Why generic FAQ pages get ignored twice
There are two separate mechanisms punishing a generic FAQ page, and fixing one does not help with the other.
In ordinary search, you are competing on a question that national sites with enormous authority already answer well. A page explaining what escrow means on an individual agent's site is not going to outrank the large financial and property sites doing the same thing. Nothing about your page is worse. It is simply the weakest entrant in a contest that was decided years ago.
In AI answers, you lose for a different reason. There is no contest at all, because no retrieval step ever happens. The model answers and moves on.
Local specificity breaks both problems at once. A national site cannot write about closing timelines in your city with any authority, and a model cannot answer it from memory. That is the same gap from two directions, which is why it is the only part of an FAQ page worth real effort. The broader mechanics of being quotable are covered in our post on what AEO means for real estate.
What actually happened to Google's FAQ rich results
A lot of advice still circulating tells agents to add FAQ code so they get expandable dropdown questions under their result in Google. That reason no longer exists, and it is worth being precise about the sequence.
In August 2023 Google changed how FAQ rich results were shown, restricting them to well-known, authoritative government and health websites. Ordinary sites stopped seeing the feature regularly from that point.
Then Google went further. Its FAQPage structured data documentation states that the feature will no longer appear in Google Search starting May 7, 2026, and that the FAQ search appearance, the rich result report, and support in the Rich Results Test are all being dropped.
Two practical consequences. Anyone selling you FAQ markup on the promise of those dropdowns is selling something that does not exist. And if someone told you your FAQ page failed because the markup was wrong, the markup was probably fine.
So is structured data still worth adding
First, in plain words, what structured data does. It is extra code on a page, invisible to visitors, that labels the parts of the page so a machine does not have to guess. On an FAQ page it says: this piece of text is a question, and this piece of text is its answer. Without it, a machine sees a heading followed by a paragraph and infers the relationship.
Given that Google's FAQ feature is going away, the honest position is this. Adding FAQ markup is worth it when it is free, and not worth a project.
Free means your website generates it automatically from your content, which is how it should work. Leave it on. The labelling still helps any system parsing the page, and Google is only retiring one search feature rather than declaring the markup harmful.
Not worth a project means do not pay someone to hand-write question and answer code for 30 questions, and do not rebuild your FAQ page to accommodate it. That budget belongs in the questions themselves.
Other structured data on an agent site has a stronger case, particularly the markup describing you and your business, because that is what machines read to work out who you are. Our post on schema markup for real estate websites covers which types earn their keep.
Phrasing: write the question the way people ask it
An FAQ question has to match how a real person asks, including the parts that are technically imprecise.
People search "how much do I need saved to buy a house", not "what are the capital requirements for residential acquisition". They ask "can I back out after making an offer", not "what are the conditions for contractual rescission". Write the human version as the question and put the correct professional term in the answer where accuracy matters.
Questions asked out loud to an assistant run longer and more conversational than typed searches. "What should I know before buying a condo with a special assessment coming" is a plausible spoken question and an implausible typed one. Both are worth having.
Where do you get the real phrasing? Your own message history. Go through six months of client emails and texts and write down every question asked more than once, in the client's words. That list is better than any keyword tool, because it is evidence rather than estimation. Cross-check it against the People Also Ask box in Google and against the queries showing up in your search console data, and write the ones that appear in all three.
Answer structure: the answer first, then the explanation
Most agents write FAQ answers backwards. They give context, then reasoning, then finally the answer in the last sentence. That is how a careful professional explains something in person, and it is the wrong shape here.
Put the direct answer in the first sentence or two. Then explain, add the exception, and give the local detail.
The reason is mechanical. A system extracting a passage from your page takes the opening of the answer. If your first sentence is "That depends on several factors", that is what gets quoted, and it is useless to everyone. If your first sentence is "Usually four to six weeks, with financing being the step most likely to add time", the quote is worth reading and worth attributing.
Human readers behave the same way. They scan for the answer and only read the explanation if the answer applies to them.
Three to five sentences is the range that works for most questions. Shorter tends to be too thin to be useful. Much longer means you are writing an article, and an article should be its own page with a link from the FAQ.
Two habits that raise the quality of every answer. Include a real number, date, or local detail wherever one exists, because that is the part nobody else can write. And put a visible last-reviewed date on any answer that contains a number, then check those twice a year, because a confidently wrong answer costs more trust than a missing one.
One FAQ page, or FAQ blocks on every page
Most agents default to one FAQ page because it is tidy. Per-page blocks usually work better.
The reason is context. Three questions about buying in a specific neighbourhood, sitting on that neighbourhood page, reinforce what that page is about and answer a reader at the moment the question occurs to them. The same three questions on a general FAQ page sit between a question about commission and a question about staging, matching nothing in particular.
A general FAQ page still has a job. Practical questions about working with you belong there: how you communicate, what you charge, what happens at a first meeting, which areas you cover. Those are about you rather than about a topic, and they have no other natural home.
So the arrangement that works for most agent sites is a short general page about working with you, plus targeted question blocks on the pages that already cover a topic. Keep every question in exactly one place. The same question answered twice on two pages competes with itself and gives a machine two candidates for one answer.
The takeaway
The only FAQ questions worth writing are the ones an AI assistant cannot answer without looking something up, which in practice means local, current, or situational questions from your own message history. Lead every answer with the answer, put a date on anything with a number in it, and put the questions on the pages they belong to rather than in one pile. The dropdown results that FAQ markup used to earn are gone from Google as of May 2026, so the page has to justify itself on being genuinely quotable now, which is a higher bar and a more durable one.
