# Martenfield AI visibility and citation for proptech and commercial real estate software. We measure whether AI answer engines cite your company, then publish the data layer that makes them. For proptech and commercial real estate software. Site: https://martenfield.com Method version: v1.0 Engines measured: ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews Runs per query: 3 What we do not promise: Pipeline numbers from AI answers; Demo counts attributed to a model; Attributed revenue from generated answers. Attribution from generated answers is unreliable in 2026. Citation share is the leading indicator we contract on, not a revenue figure. Who we turn down: Individual agents and local brokerages; No proprietary data to publish; Budget under $3,000 a month; First question is the monthly rate. ======================================================================== # SERVICES ======================================================================== ## Audit, $3,500 (one-time), 2 weeks URL: https://martenfield.com/services/audit/ We measure where you stand across five engines, map the sources they draw from, and hand you a ranked 90-day roadmap. Who this is for: Proptech and CRE software companies who suspect they are absent from AI answers and want to know the size of the problem before committing budget to it. What you get: - 50 buyer-intent queries across 5 engines at 3 runs each - Source map: which domains the models cite for your category - Entity check: how the models describe you, and what they get wrong - Hallucination log with the exact prompts that produced each error - Technical review: crawler access, schema, rendering, freshness - 90-day roadmap ordered by effort against expected movement How it runs: 1. Query set agreed. We draft 50 buyer-intent queries from your category, your competitors, and the questions your sales team already answers. You approve the set before anything runs, because the queries are the contract. 2. Measurement. Every query runs against five engines, three times each, logged out. Raw responses are stored, not just the parsed result, so a parser correction later re-reads history instead of re-running it. 3. Source and entity analysis. We extract every cited domain into a ranked source map, then check how the models describe you against how they describe the competitors they name instead. 4. Debrief. A working call on the findings and the roadmap. This is also where we tell you if the honest answer is that you do not need us yet. Not included: - Implementation. The audit tells you what to fix, it does not fix it. - Content production. - Ongoing tracking after the two weeks. Questions: Q: What do I actually get at the end? A: A source map of the domains the models cite for your category, a hallucination log with the exact prompts that produced each error, a technical review covering crawler access and schema, and a 90-day roadmap ordered by effort against expected movement. Plus the raw measurement data. Q: Can you guarantee we will get cited? A: No. We contract on measuring citation share identically before and after, and on doing the work the roadmap specifies. Anyone guaranteeing a citation outcome is guessing, because the engines change ranking behaviour without notice. Q: Why five engines and three runs? A: Generated answers vary between runs. One run tells you what a model said once. Three runs across five engines tells you what it tends to say, which is the thing you can act on and the thing that can be re-measured honestly in 90 days. Q: Do you work outside proptech and CRE? A: Not currently. The evidence, the query packs, and the source maps compound inside one category and do not transfer cleanly out of it. We would rather say no than sell you a generic version. ## Build, $12,000 to $18,000 (project), 6 to 8 weeks URL: https://martenfield.com/services/build/ We do the work. A publishing pipeline from your database, source-grade assets, and placements where the source map points. Who this is for: Companies holding proprietary data that their buyers ask about and their market cannot currently read. Usually the audit came back showing the source map is dominated by review sites and trade press, and none of it is yours. What you get: - Data publishing pipeline from your existing database - 8 to 12 source-grade assets built from data you already hold - Placements where the source map points, not where it is easy - Schema and entity architecture across the published layer - Crawler access corrected and verified against live requests - Baseline and 90-day re-measurement on the same query set How it runs: 1. Baseline. The same measurement the audit runs, on the same query set, before any work starts. Without it the 90-day number means nothing. 2. Technical layer. Crawler access, rendering, schema, and entity architecture, verified against live crawler requests rather than against a testing tool. 3. Publishing pipeline. We build the path from your database to public, readable pages. This is the part that compounds, because it keeps producing assets after we leave. 4. Source presence. Profiles claimed and corrected, data contributed to industry reports, methodology placed where the source map says the models actually look. 5. Re-measurement. Same queries, same engines, same run count, 90 days later. We publish the delta, including when it is smaller than we hoped. Not included: - Paid media or ad management. - Full-funnel content marketing. We publish source-grade assets, not blog volume. - Ongoing tracking after the engagement, unless it continues as a retainer. Questions: Q: Why is the price a range? A: The technical layer and the publishing pipeline vary with the state of your stack and how your data is stored. We quote a fixed number inside that range after the audit, and we cut scope rather than rate if the budget is tight. Q: Do we need the audit first? A: Effectively yes. The Build works from a source map and a baseline, and both come out of the audit. Buying the Build without them means paying us to guess at the order of work. Q: What is a source-grade asset? A: A page a model will cite as evidence rather than skim as marketing: a defined method, dated data, a stated sample, and figures a reader could check. Usually built from data you already hold and have never published. Q: Who owns what you build? A: You do. The pipeline runs on your infrastructure and the pages sit on your domain. There is no dependency designed into the engagement. ## Retainer, From $4,500 (per month), 6-month minimum URL: https://martenfield.com/services/retainer/ Weekly tracking, continuous publishing, and source acquisition. For companies treating citation share as a standing metric. Who this is for: Companies treating citation share as a standing metric rather than a project. Usually this follows a Build, once the pipeline exists and the question becomes keeping the position rather than taking it. What you get: - Weekly tracking across the contracted query set - 3 to 4 source-grade assets per month - Source acquisition: profiles, data contributions, trade press - Monthly report with the movement and what caused it - Quarterly benchmark against named competitors How it runs: 1. Weekly tracking. The contracted query set runs weekly. Citation sets shift when engines update, so monthly tracking finds out too late to react. 2. Continuous publishing. Three to four source-grade assets a month, chosen from where the source map moved rather than from a content calendar agreed in advance. 3. Source acquisition. Legitimate presence in the ecosystems the models trust. Profiles corrected, data contributed, methodology cited by trade press, answers given where practitioners actually ask. 4. Monthly report. The movement, and what caused it. Where we cannot attribute a change to our work, we say so. Not included: - A dashboard login. The report is a document, because that is what gets read and forwarded. - Guaranteed monthly citation growth. - Work outside the contracted query set without re-scoping. Questions: Q: Why a six-month minimum? A: Source presence is the part that moves the number most, and it is slow. Profiles, data contributions, and trade press placements take weeks to be crawled and longer to be trusted. A three-month retainer would bill you for the setup and end before the payoff. Q: Can we start with the retainer instead of the audit? A: We would rather you did not. Without a baseline we cannot show you what changed, which means you are paying monthly for work you have to take on faith. Q: What happens if the numbers do not move? A: We show you that they did not, and why we think so. Model behaviour changes and some categories are harder than others. The tracking is honest in both directions or it is not worth running. ======================================================================== # GUIDES ======================================================================== ## AI visibility tools compared URL: https://martenfield.com/guides/ai-visibility-tools-compared/ Published: 2026-07-31 What Profound, Peec, Otterly, Scrunch, and the rest of the AI visibility tracking category actually measure, and the questions to ask before you buy one. Most of this category launched inside eighteen months, raised heavily, and now sells the same headline: find out whether AI mentions your brand. The differences that matter are underneath that sentence. This is a functional comparison. We have deliberately not published a pricing table, because published prices in this category change faster than a guide can be maintained, and a stale price is worse than no price. Check current pricing on each vendor's own page before you shortlist. ## The four things these tools actually do Every product in the category is some combination of four jobs. Almost none of them do all four well, and the marketing rarely tells you which one you are buying. **Prompt running.** Sending a set of questions to one or more engines on a schedule. The differentiators are how many engines, how many runs per prompt, and whether you control the prompt set or accept a generated one. **Answer parsing.** Turning a generated answer into structured records: which brands were named, in what order, with what sentiment, and which URLs were cited. This is where the quality difference lives, and it is the hardest part to evaluate from a demo. **Change tracking.** Keeping the history so you can show that something moved. Sounds trivial. It is the reason most spreadsheet approaches collapse in month three. **Recommendation.** Telling you what to do about it. Treat this part with suspicion across every vendor in the category, including the ones with the largest funding rounds. Nobody has a validated causal model linking a specific site change to a specific citation outcome, because the engines do not publish ranking behaviour and change it without notice. ## The questions that separate them Ask these of any vendor, including us. The answers are more diagnostic than any feature list. ### How many runs per prompt, and do you report the variance? Generated answers are not deterministic. Ask the same model the same question three times and you will frequently get three different vendor lists. A tool reporting one run per prompt is reporting a coin flip and calling it a metric. If a vendor cannot tell you their run count, they are either running once or they have not thought about it. Both are disqualifying if you intend to contract on the number. ### API or consumer interface? Most tools query models through APIs, because it is cheaper, faster, and scriptable. Your buyers use the consumer apps, which have different retrieval behaviour, different system prompts, personalisation, and often a different model version. The gap is real and it is not always small. There is no fully clean answer here, since scraping consumer interfaces at volume is fragile and against most terms of service. What matters is that the vendor tells you which one they are measuring, rather than letting you assume it is the one your buyer uses. ### Do you store raw responses, or only the parsed result? This is the question that reveals whether a vendor expects to be doing this in two years. Parsers are wrong sometimes. Brand names collide with common words, competitors get missed, a citation gets attributed to the wrong sentence. If the vendor stored only the parsed output, a parser fix cannot be applied backwards, and your history is permanently built on the old mistake. If they stored the raw answer, they re-parse and your history corrects itself. ### Which engines, and does the list include Google AI Overviews? Most tools cover ChatGPT, Perplexity, and Gemini. Coverage of Google AI Overviews is the common gap, because it needs a search results provider rather than a model API, and vendor claims about the fidelity of that data conflict with each other. Ask specifically how AI Overviews data is obtained. If the answer is vague, the coverage is probably thin. ### Can I export the underlying rows? You want one row per run, per query, per engine, per brand, with the mention type, the cited URL, the position, and the timestamp. If a tool only exports its own charts, you cannot audit it, and you cannot take your history with you when you leave. ## Where the named tools sit Broad strokes, and check each vendor's current documentation before relying on this, because the category is moving quickly. **Profound** is the enterprise end. Heavily funded, broad engine coverage, built for large brands with a team to run it. The depth is real and so is the price floor. **Peec** grew fast in the mid-market on a cleaner, more focused product. Strongest when you have a defined brand and competitor set and want the tracking to be straightforward rather than exhaustive. **Otterly** is the low-cost entry point. Genuinely useful for answering "am I mentioned at all", and honest about being a monitoring tool rather than a strategy platform. **Scrunch** sits toward the enterprise end with an emphasis on brand presence across answers rather than pure citation counting. The category also has a long tail of tools that are a scheduled prompt runner plus a dashboard. There is nothing wrong with that, and it is worth knowing that is what you are paying for. ## What none of them can tell you No tool in this category can tell you what a citation is worth to you. Attribution from generated answers is unreliable. Referral data is incomplete, many answers produce no click at all, and the buyer who reads your name in an answer and searches for you directly the following week arrives as organic brand search. Any vendor quoting you pipeline attributed to AI answers is modelling, not measuring. That is not an argument against measuring citation share. It is an argument for knowing what the number is: a leading indicator of whether you are in the consideration set when a buyer asks a model, which is a real thing to care about and a different thing from revenue. ## If you are choosing today Write your own prompt set first, before you look at any product. Twenty to fifty questions your buyers actually ask, in their words, including the ones where you would expect a competitor to win. That list is the asset. The tool is interchangeable, and most of them will happily generate a prompt set for you that measures the questions you are already winning. Then run it manually once, across the engines you care about, three runs each, logged out. It takes an afternoon. You will learn more from reading twenty raw answers than from any dashboard, and you will know what you are asking a vendor to automate. ### Questions Q: Do I need an AI visibility tool at all? A: If you only want to know whether you are mentioned, you can answer that yourself in an afternoon with a spreadsheet and a query list. A tool earns its cost when you need the same measurement repeated on a schedule, across several engines, with the history kept so you can prove something changed. Q: What is the difference between mention tracking and citation tracking? A: A mention is the model naming you in the text of an answer. A citation is the model linking a source it drew from. They move independently. You can be mentioned constantly without ever being cited, which tells you the model knows of you from training data but is not currently retrieving you. Q: Why do two tools give different numbers for the same brand? A: Because they are asking different questions, at different times, from different locations, with different prompt sets, and often through different interfaces. Generated answers vary between runs. Any tool reporting a single figure without stating run count, date, and prompt set is reporting an anecdote. ## What GEO costs in 2026 URL: https://martenfield.com/guides/what-geo-costs/ Published: 2026-07-31 Real price ranges for AI visibility and generative engine optimisation work, what sits inside each band, and why anything under $3,000 a month is usually SEO with a new label. Nobody in this category publishes prices, which is why you are reading this. We publish ours, so it costs us nothing to publish the ranges around them. ## The observed market, mid-2026 Services pricing in AI visibility and generative engine optimisation runs from roughly $1,500 to $50,000 a month. That range is close to useless as stated, because the two ends are not the same product. The credible mid-market band is $3,000 to $10,000 a month. Below that you are almost always buying content production with AI visibility language layered on top. Above it you are buying either enterprise scale, multi-market coverage, or a named practitioner's time. One-off audits sit at $2,500 to $7,500 depending on query volume and engine coverage. Fixed-scope build projects run $10,000 to $25,000. Treat these as observed ranges rather than as a survey. Pricing in this category is soft, moves quickly, and is negotiated more often than it is published. ## What sits inside each band ### Under $3,000 a month Realistically: a prompt-tracking tool subscription, a monthly report generated from it, and two to four blog posts. Practitioners now write publicly that sub-$3,000 work in this category is SEO relabelled, and that is broadly fair. This is not automatically a bad purchase. If you have no measurement at all and no in-house capacity, a cheap monitoring subscription plus someone competent writing is better than nothing. Just know that you are buying content, and price it against content agencies rather than against specialists. ### $3,000 to $10,000 a month The band where the work changes shape. You should expect measurement across multiple engines with a stated run count, a source map rather than a mention count, technical remediation, and source presence work: profiles claimed and corrected, data contributed to industry reports, placements pursued where the source map says the models actually look. The thing that distinguishes this band is that someone is doing outbound work on your behalf into third-party sources. That is slow, unglamorous, and it is where most of the movement comes from. ### Above $10,000 a month Multi-market or multi-language coverage, large query sets, a category where the competitive set is genuinely contested, or an engagement including engineering work on your own data infrastructure. Also where you find agencies charging enterprise rates for mid-market work, so ask what specifically justifies the step up. ## Why the cheap end is cheap The measurement is not expensive. This is the part the category does not advertise. Running fifty queries across five engines at three runs each is roughly 750 to 1,000 API calls. At current grounded-search pricing that is tens of dollars, not thousands. A full index measuring forty companies across a category runs in the same order of magnitude. So when a tool charges $500 a month for tracking, it is not the API bill you are covering. When an agency charges $8,000 a month, the measurement is a rounding error inside it. The cost is analysis, remediation, and the slow work of getting into sources you do not control. Anyone pricing on the basis that the measurement is expensive either has not built it or is hoping you have not checked. ## What you should refuse to pay for **Pipeline guarantees from AI answers.** Attribution from generated answers is unreliable in 2026. A vendor promising attributed revenue is modelling and calling it measurement. **Ranking guarantees.** The engines change retrieval behaviour without notice and without a changelog. Nobody can guarantee a position they do not control. **A dashboard as the deliverable.** A login with charts in it is the cheapest part of this to build. If the monthly fee is mostly access to an interface, buy the tool directly and keep the difference. **Volume content.** Twelve posts a month is a content retainer. It may be worth buying, but it is not what moves citation share, and the source map will tell you that within one measurement cycle. ## Our published prices For reference, since price anchoring should happen before a call rather than during one. | Engagement | Price | Length | | --- | --- | --- | | Audit | $3,500 | 2 weeks | | Build | $12,000 to $18,000 | 6 to 8 weeks | | Retainer | From $4,500 a month | 6-month minimum | We hold to one rule on price: cut scope rather than rate. A discounted rate for the same scope tells the next client the first number was invented. ## How to buy without wasting the first quarter Start with a fixed-scope audit, from anyone, including someone other than us. It gives you a baseline, and without a baseline no retainer can prove it did anything. Insist the audit hands over the raw measurement data, not only the report. One row per run, per query, per engine, per brand, with the cited URL and timestamp. If a vendor will not hand that over, they are keeping you dependent on their interpretation of it. Then contract the ongoing work on a named query set with a re-measurement date, measured identically both times. That single sentence in a contract does more for your outcome than any amount of diligence on the vendor's methodology page. ### Questions Q: What is a reasonable budget to start with? A: A one-off audit in the $3,000 to $5,000 range is the cheapest way to find out whether you have a problem worth spending on. Committing to a retainer before you have a baseline means paying monthly for work you cannot evaluate. Q: Why is monthly work more expensive than a one-off audit? A: Because the expensive part is not the measurement, it is source presence. Getting listed, corrected, and cited in the places models already trust takes weeks per placement and does not compress. Measurement is the cheap part and always was. Q: Is GEO just SEO with a new name? A: Partly, and anyone claiming otherwise is selling. Technical crawlability, structured data, and topical authority all still matter and are all borrowed directly from search. What is genuinely new is source presence work, meaning presence in the specific third-party sources a model retrieves for your category, which classic SEO never had to care about directly. ## How to get cited by ChatGPT URL: https://martenfield.com/guides/how-to-get-cited-by-chatgpt/ Published: 2026-07-31 The mechanism behind citations in generated answers, and the work that actually moves it. Written for people who have already read the listicle version. The listicle version of this article says: write good content, add schema, get mentioned on other sites. That is not wrong, and it is not specific enough to act on. Here is the mechanism, and then the work that follows from it. ## What is actually happening when you get cited When a model answers a question that needs current information, it does not recall your page from training. It runs a retrieval step: a search, against an index, using a query the model wrote itself from your question. It reads some of what comes back, and it writes an answer citing the pages it used. Four separate things have to go right, and they fail independently. **The retrieval query has to be one you can win.** The model rewrites "which lease abstraction tool should we use" into one or more search queries of its own choosing. You are not optimising for the user's question, you are optimising for the queries the model generates from it, which are usually more literal and more category-shaped than the original. **Your page has to be in the index behind that retrieval.** For most engines that means a conventional search index. Classic ranking did not stop mattering, it changed what it pays out. **The crawler has to be able to fetch you when it looks.** This is where a surprising number of companies fail, silently. **The content has to be quotable.** A model writing a two-hundred-word answer needs a specific, extractable claim with something that looks like evidence attached. Marketing prose gives it nothing to lift. ## Start with the failure that costs nothing to fix Check whether the AI crawlers can reach you at all. The bots that matter are separate from Googlebot and are frequently blocked by default. `GPTBot` and `OAI-SearchBot` for OpenAI, `ClaudeBot` and `Claude-SearchBot` for Anthropic, `PerplexityBot`, `Google-Extended`, and `Applebot-Extended`. Two things block them routinely. The first is a `robots.txt` written before these agents existed, often with a blanket disallow inherited from a template. The second is a bot-management product. Several CDNs and WAFs ship managed rules that block AI crawlers by default, and a growing number treat blocking them as a selling point. The failure mode is nasty because it is invisible from your side. Your own testing looks fine. Analytics show nothing wrong. The crawler simply gets a 403 and you never appear. Check it from outside with the actual user agent, not by reading your configuration. Then check your CDN's bot settings separately, because `robots.txt` permitting a crawler does not mean your edge is letting it through. ## Then make yourself worth quoting A model composing an answer is choosing sentences it can lift with a citation attached. What gets lifted has a shape. **Specific figures with a stated basis.** A lease abstraction platform stating a median tenant improvement allowance, the number of leases it was drawn from, and the quarter it covers is quotable. "We help you understand TI allowances" is not. We have deliberately not written a worked example with numbers in it here. Every page on this site is also served as plain markdown and concatenated into `llms-full.txt`, so an invented figure inside quotation marks stops being obviously hypothetical the moment something reads the plain-text version. On a site whose subject is sentences models lift out of context, that is not a risk worth taking to illustrate a point this paragraph already makes. **A stated method.** Sourced, dated, with the sample size named. This is the difference between a page that reads as evidence and a page that reads as a claim. **Direct answers near the top.** The question restated, then answered in one or two sentences, then the supporting detail. Not a five-paragraph preamble. **Dates in the visible text.** Not only in the schema. When the data was collected, and when the page was last updated. Retrieval systems and readers both discount undated material, and the model frequently repeats the date in its answer if you gave it one. **Real tables.** Marked up as tables, not as styled divs, and not rendered by JavaScript after load. ## The part most people skip Fixing your own site is necessary and usually insufficient. Run the queries you care about and write down every source that gets cited. That list is your source map, and it is almost never your competitors' websites. It is review sites, trade press, industry associations, forums, and occasionally one unusually thorough blog post from 2023. Your competitors are not beating you because their homepage is better. They are being cited because they appear in the sources the model retrieves and you do not. So the work is: claim and correct your profiles on the review sites in your source map, contribute data to the industry reports that get cited, get your methodology referenced by the trade press that covers your category, and answer questions where practitioners actually ask them. Slow, unglamorous, and it is where the movement comes from. ## Publish the data you already have The highest-leverage asset most companies own is a dataset they have never published, sitting in their production database. Lease comps. Cost benchmarks. Absorption figures. Permit timelines by jurisdiction. Whatever the operational core of your product produces. Your buyers ask models questions about exactly this, and the models answer from whatever public source they can find, which is currently somebody else. Publishing an aggregated, anonymised, dated slice of it, with a method note, is the single strongest move available to most companies in this position. It is also the one competitors cannot copy, because they do not have your data. ## What to measure Pick twenty to fifty queries your buyers actually ask. Run them across the engines you care about, three times each, logged out, and record: which companies were named, in what order, and which URLs were cited. Repeat monthly. The number that matters is the share of those answers in which you are cited, and the shape of the source map behind them. Do not measure whether you rank for "your brand name". You will always win that and it will tell you nothing. ## The honest caveat Nobody has a validated causal model here. The engines do not publish retrieval behaviour, they change it without notice, and a change in your numbers may be a model update rather than anything you did. What you can do is remove the blockers, publish material worth quoting, get into the sources that are already being retrieved, and measure consistently enough to tell the difference between a trend and a fluctuation. That is the whole discipline. Anyone selling you more certainty than that is selling. ### Questions Q: How long does it take to see a change? A: Technical fixes can show up within days to a few weeks once the crawler returns. Source presence work takes longer, typically six to twelve weeks, because a new profile or placement has to be crawled and then has to become one of the pages the retrieval layer prefers. Anyone promising movement in a fortnight is describing the technical fixes only. Q: Does publishing more content help? A: Volume on its own does not. What helps is publishing something that is genuinely the best available source for a specific question, with a stated method and checkable figures. One page like that outperforms thirty pages of category commentary, because retrieval is picking a source to quote rather than counting how much you have written. Q: Can I pay to be cited? A: Not directly, and be suspicious of anyone implying otherwise. What you can do is legitimately appear in the third-party sources the models already retrieve for your category, which is slower and is the actual work. ## Does ChatGPT read schema URL: https://martenfield.com/guides/does-chatgpt-read-schema/ Published: 2026-07-31 An honest answer to the question technical buyers ask first. Structured data is not read by the model directly, and it still matters. Here is why both are true. The short answer: not the way the question usually means, and it still matters. That sounds like a dodge. It is not, and the distinction is the whole point. ## What actually happens to your JSON-LD When a model answers a question using live information, it does not fetch your HTML, parse the `