Top 10 AI Agents forHotels and Resorts in 2026

Every ranking of hotel AI you will read this year is built on numbers nobody outside the vendor has ever checked. Open five of them and the shape repeats: a product automates 97% of guest messages, deflects 80% without handoff, delivers 23x ROI, triples conversion against the website. Those are printed as though they were measurements. They are not measurements. They are marketing assertions about the seller, published by the seller, with no methodology, no denominator, and no published definition of the word being measured.

The claim under test in this article is narrow: can an independent hotel owner rely on any published performance figure in this category?After opening every vendor's own pricing, product, integration and trust pages on 23 August 2026, our answer is no — and the useful consequence is that you should stop trying. Rank on what you can verify from a browser and a contract, put the performance question into your RFP as a written commitment rather than a number you read on a homepage, and measure the thing yourself, at your own property, with your own guests.

That is not a counsel of despair. It turns out that when you score only checkable attributes, the table separates sharply — and it separates along lines that no competing listicle reports. Three vendors publish a price; eleven do not. One vendor publishes a human-in-the-loop boundary; thirteen do not. Two vendors publish anything at all about the European transparency regulation that started binding their product three weeks before this was published. One vendor rebranded in January and is still listed under its dead name across most of the internet.

There is a second reason to distrust the figures, and it comes from a regulator rather than from us. In January 2025 the Securities and Exchange Commission entered a settled order against a drive-thru voice AI company over exactly this kind of number, finding that the product lacked the capability to take orders on its own and required substantial human involvement, and that people abroad processed the vast majority of orders. That order is discussed in detail in the restaurant edition of this series — see our ranking of the top AI agents for restaurants in 2026 — and the framing there matters: the company consented without admitting or denying the findings, the remedy was a cease-and-desist order with no civil penalty, and the findings run against that company only. It is not evidence that any hotel vendor is doing the same thing. It is evidence of what happened the one time anyone with subpoena power actually audited a published AI performance metric in hospitality-adjacent technology.

This piece sits inside a wider set of vertical rankings built on the same method; the cross-industry version, with the same refusals applied to general-purpose tools, is our guide to the top AI agents in 2026. What follows is the hotel edition: what we scored, what we would not score, ten vendors in order, four constraints that decide whether any of it works, and the parts we could not verify.

We scored eight attributes, every one of which you can re-check from a browser in under an hour. That constraint is the entire method. If an owner cannot independently confirm a cell in our table, the cell does not belong in a ranking — it belongs in a sales deck.

We refused to score accuracy, deflection, containment, resolution rate, ROI, hours saved, RevPAR lift and conversion.Not because they do not matter — operationally they are the only things that matter — but because every published figure in this market is self-reported by the seller and there is no neutral party checking any of them. Printing one vendor's deflection rate beside another's in a comparison table implies a common measurement standard. There is no common standard, and there is not even a common definition: one vendor's automated conversation may be a conversation a guest abandoned.

Two refusals are worth naming individually, because they are the cleanest illustrations available. A large Brazilian vendor publishes a 23x average ROI figure on its homepage with no methodology, no sample and no definition of return. And a major PMS publishes, on the same pricing page that declines to publish a price, a claim about payback in as little as seven months. A payback period whose denominator the vendor will not disclose is not a number. It is a marketing asset shaped like a number, and no buyer can compute it.

We also refuse the third-party aggregator prices that circulate for this category. Software directories carry figures for several vendors here, quoted per room, per user or per year. Those are not vendor-published prices, they are not attributed to a source document, and printing one would launder a directory's guess into a fact. Where a vendor declines to publish, the honest cell reads not publicly disclosed, and we never estimate.

On benchmarks, stated precisely. No independent benchmark of these hotel products exists — no head-to-head evaluation of Canary, Duve, HiJiffy, Asksuite, Akia, chatlyn or any other vendor here has been published by a party that is not selling one of them. That is a narrow claim and we hold it narrowly. The nearest thing to independent measurement in adjacent hospitality is the Intouch Insight study released with QSR Magazine on 1 October 2025, which mystery-shopped quick-service drive-thru lanes and found AI-enabled orders 83% accurate against an 87% study average, reaching 95% only when staff intervened. That study measures drive-thru ordering in restaurants, not hotel guest messaging, and it identifies brands rather than vendors. We label the transfer explicitly rather than borrowing its authority. It is not a hotel benchmark; it is simply the only place anyone has measured this modality against a human baseline, and the result points the opposite way from the marketing.

On 16 June 2026 Oracle announced new AI capabilities inside OPERA Cloud, and its own release states that all these capabilities are available today at no additional cost to OPERA Cloud customers worldwide. Most rankings of this category published since then do not mention it, which is a reasonable proxy for how carefully they were built. It changes the competitive picture, and it changes it in a specific direction that an independent operator needs to read correctly.

What shipped is the Oracle OPERA Cloud Assistant: natural-language access to operational knowledge, hotel procedures and Oracle documentation; AI-assisted room assignment drawing on reservation details, guest preferences and stay history; AI-generated rate code descriptions; and AI translation of configuration and operational content, supporting operations across 230 countries and territories. Oracle's framing throughout is about empowering associates.

The second reading is the one that applies to most people holding this article. OPERA is the enterprise-brand property management system — Oracle's release notes that Wyndham has more than 2,100 properties running on OPERA Cloud. An independent hotel or a small group is far more likely to be on Mews, Cloudbeds, Apaleo, StayNTouch, RoomRaccoon or Little Hotelier, in which case none of this reaches you at no cost or at any cost. Do not let the headline imply otherwise.

One more note on quotation discipline, since this is a market where vendor language gets reprinted as fact. Oracle's release also carries productivity and efficiency language — faster decisions, increased productivity. That is marketing, and we quote the capability rather than the benefit. The cost sentence is a capability statement and we quote it; the productivity sentences are claims and we do not.

Every cell below is either a citation you can open or the words “not publicly disclosed.”Nothing here is estimated, inferred from a directory, or taken from a vendor's claim about a competitor. Where a vendor's page returned an error, the cell records the error, because a 404 on a pricing page is itself a finding about how the vendor sells.

Read the pricing column first, because it is the column that most changes a small operator's day. Three of the fourteen vendor names we checked publish a price at all. HiJiffy publishes tier floors in euros, dollars and pounds, with setup fees stated per five properties. Duve publishes minimum package amounts. Visito publishes a credit-metered ladder — and Visito is the vendor we could not rank, for reasons in the next section. Everyone else asks you to book a call before you can find out whether the product is in your budget range at all.

Read the compliance column second, and read it for specificity rather than for logos. There is a real difference between a trust page that says PCI DSS and one that says PCI DSS v4.0.1 and publishes a shared responsibility matrix. Only one vendor here does the latter. And be alert to a trap this method has caught repeatedly in other verticals: a vendor citing a certification that belongs to its cloud hosting provider rather than to its own legal entity. Before you credit any compliance cell, confirm the certificate names the vendor.

The order below reflects how much a buyer can verify before signing, not how good the AI is. We have no way to score how good the AI is, and neither does anyone else writing a ranking of this category. Each entry states what the product does, what is verifiable, what is not disclosed, who it fits, who it does not, and who owns it.

Considered and not ranked: four names appear on most competing lists and are not in our table, and the reasons are worth as much as the table. A vendor we cannot confirm still trades in 2026, under a name we can attach to a legal entity, does not get ranked — because the first thing a ranking owes a buyer is that the companies in it exist as described.

The thing that decides whether any of this works is not the model. It is that a hotel AI agent quoting a rate is making your offer, and there is now a federal rule and a stricter California statute governing what that offer must contain. No vendor safe harbour exists. The FTC would name the business that offered the room, not the chat vendor that rendered the sentence.

The Trade Regulation Rule on Unfair or Deceptive Fees, 16 CFR Part 464, was published at 90 FR 2166 on 10 January 2025 and took effect on 12 May 2025. Section 464.1 defines a covered good or service to include short-term lodging — temporary sleeping accommodations at a hotel, motel, inn, short-term rental, vacation rental or other place of lodging. It defines total price as the maximum total of all fees or charges a consumer must pay, plus any mandatory ancillary good or service, except that government charges, shipping charges, and fees for optional ancillary items may be excluded.

Section 464.2(a) makes it an unfair and deceptive practice to offer, display or advertise any price of a covered good or service without clearly and conspicuously disclosing the total price. Section 464.2(b) requires the total price to be disclosed more prominently than any other pricing information. Section 464.2(c) requires disclosure, before the consumer consents to pay, of the nature, purpose and amount of any excluded fee, and of the final amount of payment. And §464.3 prohibits misrepresenting a fee's nature, purpose, amount or refundability.

There are two provisions written for exactly the technology this article is about. The clear-and-conspicuous definition at §464.1 states that in any communication using an interactive electronic medium — the internet, a mobile application, or software — the disclosure must be unavoidable. And an audible disclosure, including by telephone, must be delivered at a volume, speed and cadence sufficient for ordinary consumers to easily hear and understand it. A voice booking agent that recites a total price at machine pace is inside that sentence. So is a chat agent that puts the total behind a “see details” expander.

If your property takes European guests, the transparency obligation applied on 2 August 2026 — three weeks before this article was published — and it binds your guest-facing agent today. The most common summary in circulation right now is that the EU delayed the AI Act to 2027. That summary is wrong for the only provision that reaches a hotel chatbot, and it is the kind of wrong an operator acts on.

Here is what actually moved. The Digital Omnibus, in the European Commission's own words, entered into force on 27 July 2026, bringing extended timelines. It deferred the high-risk obligations: high-risk AI systems in Annex III now apply from 2 December 2027, and high-risk AI embedded in physical products under Annex I from 2 August 2028. Those two dates are real, and if you have been told the AI Act slipped, those are the obligations being described.

Notice why the mistaken summary spreads so easily. The Commission's own omnibus announcement does not mention Article 50 at all — it is about the obligations that moved. Read quickly, an announcement about extended timelines reads as though everything moved. Nothing about the guest-facing transparency duty moved.

There is one nuance an operator must not be talked out of. A hotel is normally a deployer of an AI system, not its provider, and the Article 50(1) design duty sits on the provider. It is therefore tempting for a vendor to say your vendor handles it. The practical consequence for you is different: the disclosure has to actually appear in the guest interaction, on every channel, and a hotel that configures it away — because it looks clumsy in the brand voice, or because someone removed it during a redesign — owns that outcome. Treat the disclosure as non-configurable and audit it quarterly by opening the widget as a guest. The engineering detail of doing this well, across chat, voice and email, is covered in our guide to engineering EU AI Act Article 50 transparency.

This is also a scoring column no competing listicle has. Three vendors in this market publish anything at all about the regulation that started binding their product three weeks ago: Canary Technologies, with an EU AI Act Transparency Compliance Statement on its trust centre; chatlyn, which states that guests are informed when AI replies; and Mews, which carries an EU AI Act entry on its trust centre. Eleven do not. That is not proof the other eleven are non-compliant — it is proof that you will have to ask them, in writing, and that three of them have already answered.

Three more constraints decide whether a hotel agent is deployable, and all three are about what the agent touches rather than what it says. Cards pull the messaging archive into payment-card scope. Voice pulls it into state recording law. Outbound messaging pulls it into the TCPA. None of these is exotic, and all three are routinely discovered after launch.

The property. Consider a 140-room independent resort handling roughly 2,400 inbound guest messages a month across web chat, WhatsApp and SMS, plus about 1,100 inbound phone calls. It sells a mandatory $32 nightly resort fee. It takes European guests in summer. It is in California.

The labour denominator, done honestly. The Bureau of Labor Statistics puts the median hotel, motel and resort desk clerk at $16.86 an hour, or $35,070 a year, in its May 2025 estimates, across 261,420 people employed nationally. That is base wage only — it excludes employer payroll taxes and any benefits. Now add the sleeper fact: average weekly hours in accommodation were 29.6 in June 2026. Accommodation is a part-time-heavy industry, so any vendor model that prices its savings against a 40-hour full-time equivalent is overstating the labour it displaces by roughly a quarter before it starts.

The message arithmetic. Suppose each of those 2,400 monthly messages consumes a median four minutes end to end — reading, checking the reservation, answering, logging. That is 160 hours a month, 1,920 hours a year, or about $32,371 of base wage at the median. If an agent genuinely handles half of them without a human touching the thread, the notional saving is 960 hours, roughly $16,186 of base wage a year.

Now the costs, using only published numbers. The lowest published subscription floor in our table that would plausibly cover this property is HiJiffy's Pro tier at €159 a month, with a setup fee stated per five properties — call it roughly $2,100 a year at the floor, before any scaling for room count. Set against a $16,186 notional saving, that looks like a landslide. It is not, and the reason is the line item vendors do not quote: integration. Connecting an agent to your property management system, your booking engine, your payment tokenisation and your knowledge base, then instrumenting it so you can measure it yourself, is our single-workflow agent band — $28k to $70k over four to nine weeks. On the low end, year one is net negative and the crossover lands early in year two.

And here is the question the crossover conceals. Hours saved only become money when a shift disappears from the schedule. In an industry averaging 29.6 weekly hours, the shift usually does not disappear — the work gets reabsorbed, and the desk clerk who was answering WhatsApp now does something else that also needed doing. That may be an excellent outcome. It is not a cost saving, and it should not be modelled as one. The honest business cases we see in this vertical are built on three things instead: response coverage at hours when nobody is at the desk, consistency of answers when the person who knew the property left last quarter, and revenue from upsells that were previously never offered. Note that the third one is the one that carries all the fee-disclosure and TCPA exposure described above.

The turnover point, stated precisely. The quits rate in accommodation and food serviceswas 4.5% in June 2026 — roughly one in twenty-two employees voluntarily leaving every month — against a job openings rate of 4.6% and total separations of 5.7%. That series is NAICS 72, which combines accommodation with food services and is dominated by restaurants; it is not hotels alone, and no hotels-only series exists. Say it that way. The reason it belongs in the business case is that it reframes the pitch honestly: the operator's problem is often not headcount cost but institutional memory walking out, which is a real argument for encoding knowledge in a system and requires no vendor statistic at all.

Two figures we will not use, and you should not either. The claim that it costs five times more to acquire a guest than to retain one is sold alongside CRM and loyalty tooling in this exact market. Its origin was traced by Ipsos Loyalty, in a published excerpt from the book Loyalty Myths, to research conducted by the Technical Assistance Research Project in Washington, D.C. in the late 1980s that nobody can now produce — with several other consultancies claiming the same finding as their own, and credibility arriving via a 1990 Harvard Business Review article about service recovery. The same document says there is enough contrary information to bury or significantly qualify the truism, and closes by warning that a retention strategy based wholly on the myth is a recipe for financial disappointment. That is the firm that published it for years, retracting it in print. The second figure is any specific percentage for how much more profitable direct bookings are. We found no primary source for any of them, and the claim is structurally unfalsifiable as usually stated because it books commission against zero while ignoring metasearch bids, brand advertising, booking-engine fees, payment processing, loyalty discounts and the labour of running the channel. A direct booking bought with a 12% metasearch cost of acquisition is not commission-free. Write the arithmetic for your own property; do not print the statistic.

The same discipline applies to OTA commission percentages. Every figure we surfaced in this market traces to a company selling hotels a way to reduce OTA commissions, which is textbook circular sourcing. If you need the number, read your own OTA contract — which is verifiable by you, and is the whole point of this method.

The first thing that breaks is usually not the agent. It is something else on the same integration gateway. This is the most useful technical fact in this vertical, it is documented by Oracle in public, and we have not seen it in a single competing ranking.

Oracle's documented limits for the Hospitality Integration Platform state that API throttling permits up to 50 requests per second per gateway, shared by all consumers and clients accessing that gateway, with a short-term burst allowance to 100 requests per second, also shared. Excess requests are delayed; beyond the burst threshold they are rejected with HTTP 429. Read the word shared twice, because it carries the whole failure mode.

Three more documented ceilings belong in any design review before you scale. Streaming API management is limited to 12 requests per minute, bursting to 100 per minute. Only one client may consume events from a single chain, in a single gateway, on a single application key — which turns two vendors both wanting the event stream into a straight architectural conflict rather than a configuration setting, and the symptom is an absence of events rather than an error. Streaming consumers must reconnect at least once every 24 hours or pass an offset, with a minimum of 10,000 milliseconds between closing and reopening a websocket. A maximum of 100 applications may be registered in the developer portal. And total HTTP header size is capped at 8 KB including the OAuth token and application key, which is small enough to matter once you are passing both. Oracle also notes it may adjust these thresholds as needed, so treat the numbers as current rather than permanent.

There is a buying heuristic hiding in this. The two property management systems with the most open public documentation — Apaleo, whose developer platform exposes separate Swagger surfaces for core, payment, identity, webhooks, distribution and automation with OAuth 2.0 grants for both multi-hotel and single-hotel apps, and Mews, whose Connector API needs no login — are the ones where an independent can actually get an agent integrated without a partnership negotiation. Oracle's OHIP is well documented but gated by partner onboarding and hard shared limits. That difference, not any vendor's feature list, is what decides whether an AI project at an independent hotel is a six-week job or a six-month one.

One failure mode deserves its own paragraph, because it has no clean fix. Prompt injection is unsolved. An agent that reads untrusted guest input — an OTA message, an inbound email, a review, an uploaded document — is reading text that an adversary may have written specifically to redirect it. There is no filter, no model and no vendor that has solved this, and any vendor telling you they handle it is telling you they have not thought about it. The only reliable posture is blast-radius reduction: read-only by default, an explicit allowlist of tools rather than a denylist, scoping every action to the current reservation, and human approval on anything that moves money, changes a booking, or sends a message outside the thread. Design as though the agent will one day be successfully instructed by a guest, and decide now what you are comfortable with it doing on that day.

Every row below traces to a constraint verified earlier in this article, not to opinion. This is the table to put in front of a vendor and ask them to sign, because with one exception nobody in this market publishes one. That exception is Alliants, which publishes a boundary by disclaiming autonomy outright.

Two rows are worth arguing about, and you should argue about them internally before a vendor does it for you. The first is the goodwill concession. At least one vendor here markets service recovery in language implying the agent decides what to offer a disappointed guest. There is no law against that. There is a blast-radius argument against it, and it is the same argument as for prompt injection: an agent authorised to give things away is an agent an adversary is motivated to talk to. Put a hard currency limit in configuration and a human on anything above it.

The second is being the only channel. It is tempting, especially overnight at a small property, to let the agent be the entire front desk between midnight and six. Do not. There is no ADA Title III web accessibility rule to comply with, which means there is no checkbox that makes an AI-only channel defensible, and the absence of a rule is worse for an operator than a rule would be. Keep a human path, publish it, and make it reachable from the first turn of the conversation.

These are our published bands, identical across every vertical we write about, and they cover the build rather than the vendor subscription. A hotel AI project has two cost lines and vendors quote one of them.

Senior-led work is $150–$225 per hour. Retainers run $2,500–$9,500 per month. There is a 30-day post-launch warranty, and a fixed-price phased proposal follows within 5 business days. Full source-code and IP ownership transfers on delivery. We are a senior-led Black-owned agency in Los Angeles, reachable on +1 (424) 272-5601, and the only booking link we use is calendly.com/frenchydigital/discovery-call.

One sequencing note that saves more money than any technical decision: do the fee-disclosure and consent work during discovery, not after the pilot succeeds. Rewriting how your agent quotes a rate costs one sprint before launch. Discovering after eight months that every California quote omitted tax before the reservation step is a different kind of problem, and §17568.6(e) prices it at up to $10,000 per violation.

Each of these is something we observed while building this ranking, not a generic warning list. Treat any two appearing together as a reason to slow the process down.

• A deflection, automation or ROI figure with no published definition. Ask what the denominator is, whether an abandoned conversation counts as contained, and whether any human — including a remote one — touches the thread.
• A payback period published on a page that carries no price. If the vendor will not disclose the denominator, the buyer cannot compute the number, and it is not a number.
• Refusal to state a human-in-the-loop authority limit in writing. Exactly one vendor in this market publishes one voluntarily. The ones that will commit contractually are telling you something the others are not.
• An integration count with no directory behind it. “400+ integrations” with no named systems and no docs URL scores as undocumented. A named list of eighteen platforms you can spot-check beats it outright.
• A compliance badge with no version. “PCI DSS” is a materially weaker claim than “PCI DSS v4.0.1” with a shared responsibility matrix, and the difference is exactly the part a buyer needs.
• A certification that belongs to the vendor's cloud host rather than to the vendor's own legal entity. Confirm the certificate names the company you are contracting with; this trap has caught vendors in several adjacent markets.
• A pricing page that returns 404, or a trust page that does not exist. Both happened in this roster. A missing page is not neutral — it tells you how the vendor expects the sale to go.
• A price quoted to you from a software directory rather than from the vendor. Aggregator figures circulate for most vendors here and none of them is a vendor-published price.
• Silence on the EU AI Act while selling you a guest-facing agent. The transparency obligations applied on 2 August 2026. Three vendors here have published a position; ask the rest for theirs in writing.
• An upsell flow that pre-selects the add-on, or discloses the charge after the total. That is the Hopper fact pattern in a conversational channel, and the Fees Rule count in that case was pleaded on short-term lodging.
• Any suggestion that the agent can take a card number in chat to make it easier for the guest. That single convenience pulls your entire messaging archive into cardholder-data scope.
• A promise that prompt injection is handled. It is not handled anywhere by anyone. The correct answer describes blast-radius controls, tool allowlists and approval gates.
• A roster or a case study that lists an acquired product under its old independent name. If a vendor's competitive deck still says Quicktext, OpenKey, Easyway, Whistle or NAVIS, the rest of that deck is the same vintage.
• Reluctance to run a limited pilot with your instrumentation rather than their dashboard. The measurement should be yours, on your property, or it is not a measurement.

This section is part of the product, not a disclaimer at the end of it. Everything below is something we tried to check on 23 August 2026 and could not, or could only check through a secondary source. Naming it is the difference between a ranking and a sales page.

• No independent benchmark of any hotel AI product exists. No head-to-head evaluation of any vendor in this table has been published by a party not selling one of them. The Intouch Insight drive-thru study we cite measures restaurants, not hotels, and identifies brands rather than vendors.
• Compliance posture is unverified for seven of the fourteen vendors we checked. Asksuite, Akia, Visito, HiJiffy, Quinta, Operto and Revinate had no trust page located or checked in this pass.
• Three vendor pages failed on the day we checked: HiJiffy's security page returned HTTP 404, Asksuite's pricing page returned HTTP 404, and the Cloudbeds PCI support article returned HTTP 403 to automated fetch. All three need a browser check before anyone relies on them.
• The Official Journal would not serve the AI Act text to any automated request in this pass. The Article 50 duty described here is a paraphrase, deliberately not presented as a quotation of the regulation, and the 2 August 2026 applicability date should be confirmed against Article 113 of the regulation itself.
• Legal entity names are not disclosed on the websites of Quinta, Akia, Operto, Visito and Asksuite. Country of incorporation is widely reported for several of them and we did not print any of it as fact.
• Akia's 2026 trading is inferred from a live site with current customer logos and a working demo booker, not proven by a dated 2026 item. That is why it does not rank higher.
• Data residency and retention commitments were not found for any vendor except Mews, which publishes hosting and a named subprocessor list, and chatlyn, which asserts EU hosting without an audit. That is a real gap: it is a standard procurement question nobody in this market answers publicly.
• No vendor publishes a human-in-the-loop authority boundary except Alliants, and Alliants does so by disclaiming autonomy entirely rather than by defining a limit.
• Resort-fee attorney-general settlements involving large chains were not verified to primary consent orders, so this article prints nothing specific about them. An assurance of voluntary compliance, a consent decree and a judgment are three different instruments and we could not confirm which applied.
• The PCI DSS requirement numbering for payment-page script inventory, tamper detection and third-party management is taken from secondary summaries. Confirm the exact wording and numbering against the PCI Security Standards Council document library before it enters a policy document.
• The DOJ interim final rule extending ADA Title II web compliance dates is described from secondary sources; we did not obtain the Federal Register citation, so no compliance dates for it appear in the body of this article. The report that the Title III rulemaking is paused indefinitely is likewise secondary.
• Reports that OHIP API usage is billed per call, and that partner certification can run long, came from a third-party API review rather than Oracle's own documentation. Both would be commercially significant for an independent, and neither is asserted here.
• The July 2026 BLS accommodation employment figure is preliminary and CES revises for two months, so treat it as provisional. Accommodation employment has been flat to slightly down across 2026 — it does not support a claim that hotels are shedding staff to AI, and we make no such claim.
• Vendor employee counts, customer counts, funding totals and valuations are as published by the vendor or reported in trade press, and we audited none of them. Where sources contradicted each other — HiJiffy's funding and deployment counts, Canary's valuation, the month of the Dack acquisition — we printed nothing.
• Cloudbeds and Revinate pricing was not verified on the vendors' own sites in this pass and is recorded as not publicly disclosed rather than as confirmed gated.