BLACK, Circuit Judge:

The issue before this Court is whether the district court correctly denied Plaintiff-Appellant Mary P. Dahl-Eimers' application for a preliminary injunction. Dahl-Eimers has advanced breast cancer, and she wants Defendant-Appellee Mutual of Omaha Life Insurance Company to pay for high dose chemotherapy with autologous bone marrow transplant (HDC-ABMT) under their major medical expense insurance contract. Mutual of Omaha refused to provide coverage for HDC-ABMT because the company considers it experimental and therefore not medically necessary. The insurance policy covers only medically necessary services, and under the policy medical services that are considered experimental are not medically necessary.

In Dahl-Eimers' application for a preliminary injunction to enjoin Mutual of Omaha from denying coverage under the policy, the district court ruled that she did not have a substantial likelihood of success on the merits. 812 F. Supp. 1193. The court based its ruling in part on its legal determination that the contract term "experimental" is not ambiguous. As explained below, we conclude that the phrase "considered experimental," standing alone in a major medical insurance policy, is ambiguous as a matter of law, and we reverse the district court's decision.

I. BACKGROUND

Dahl-Eimers has Stage IV metastatic breast cancer, which is considered incurable by traditional cancer treatments, but which may respond to chemotherapy drugs in high doses. High dose chemotherapy, combined with autologous bone marrow transplant, has recently been shown to be an effective treatment in some breast cancer cases, and Dahl-Eimers' treating physician referred her to the H. Lee Moffitt Cancer Center and Research Institute for HDC-ABMT treatment.

The insurance contract between Mutual of Omaha and Dahl-Eimers provides for reimbursement of expenses incurred for medically necessary services or supplies. A medically necessary service or supply is defined in the contract as one that:

(a) is appropriate and consistent with the diagnosis in accord with accepted standards of community practice; (b) is not considered experimental; and (c) could not have been omitted without adversely affecting the insured person's condition or the quality of medical care.

Mutual of Omaha refused to pay for HDC-ABMT treatment because it determined that the proposed treatment was not medically necessary and therefore was not covered by the policy.

Before the district court, Dahl-Eimers identified the central issues as being whether the term "experimental" is ambiguous and whether the treatment is experimental. Mutual of Omaha contended that Dahl-Eimers did not meet any of the policy's three criteria for a medically necessary service. In its order, the district court focused only on subparagraph (b) of the definition of medically necessary and determined that the critical issue was whether the term "experimental" is ambiguous.

The court held that the term "experimental," as applied to the facts of this case, is not ambiguous. It then held that HDC-ABMT treatment for breast cancer is experimental and therefore not covered by the policy. The district court concluded that Dahl-Eimers did not have a substantial likelihood of prevailing on the merits and denied her application for a preliminary injunction. The court certified that its order involved a controlling question of law as to which there is substantial ground for difference of opinion. 28 U.S.C. § 1292.

On appeal, Dahl-Eimers argues that the district court erred by focusing on the word "experimental" instead of the phrase "considered experimental." She contends that the insurance contract is ambiguous because it does not define "experimental" and it does not specify who will determine whether a service is considered experimental. She asserts that there are a number of reasonable interpretations about who will make the determination whether a particular treatment is experimental. She also argues that HDC-ABMT is no longer considered an experimental treatment for breast cancer.

Mutual of Omaha counters that the alternative interpretations offered by Dahl-Eimers are not reasonable. It recognizes that there are various potential sources which could determine whether a treatment is considered experimental, but claims that the contract would only be ambiguous if the various sources could adopt differing, yet reasonable interpretations of the term. Finally, Mutual of Omaha argues that HDC-ABMT, using the particular combination of drugs proposed for Dahl-Eimers, is experimental treatment for breast cancer.

II. STANDARD OF REVIEW

Questions of law are reviewed de novo. Salve Regina College v. Russell, ___ U.S. ___, 111 S.Ct. 1217, 113 L.Ed.2d 190 (1991). Under Florida law, interpretation of an insurance contract, including determination and resolution of ambiguity, is a matter of law. Sproles v. American States Ins. Co., 578 So.2d 482, 484 (Fla. 5th DCA 1991); Gulf Tampa Drydock Co. v. Great Atl. Ins. Co., 757 F.2d 1172, 1174 (11th Cir. 1985). The portion of the district court's denial of the preliminary injunction based solely on its legal construction of the insurance contract is likewise a question of law. See E. Remy Martin Co., S.A. v. Shaw-Ross Int'l Imports, Inc., 756 F.2d 1525, 1529 (11th Cir. 1985) (conclusions of law in action on preliminary injunction subject to broad review). The district court's findings of fact are reviewed under the clearly erroneous standard. Newell v. Prudential Ins. Co., 904 F.2d 644, 649 (11th Cir. 1990).

III. DISCUSSION

A. Applicable Law

Under Florida law, courts must construe an insurance contract in its entirety, striving to give every provision meaning and effect. Excelsior Ins. Co. v. Pomona Park Bar Package Store, 369 So.2d 938, 941 (Fla. 1979). An insurance contract is ambiguous if it is susceptible to two or more reasonable interpretations that can fairly be made. Herring v. First S. Ins. Co., 522 So.2d 1066, 1068 (Fla. 1st DCA 1988); Ideal Mut. Ins. Co. v. C.D.I. Constr., Inc., 640 F.2d 654, 657 (5th Cir. Unit B Mar. 1981). When one of these interpretations results in coverage and another results in exclusion, ambiguity exists in the insurance policy. Weldon v. All Am. Life Ins. Co., 605 So.2d 911, 915 (Fla. 2d DCA 1992); Gulf Tampa Drydock, 757 F.2d at 1174-75.

Ambiguity also may arise from silence. Cf. Davis v. Crown Life Ins. Co., 696 F.2d 1343, 1346 (11th Cir. 1983) (ambiguity resulting from silence in certificate of insurance, which did not include a controlling provision recited in the master policy, treated as an express ambiguity). Further differing interpretations of the same provision is evidence of ambiguity, Dimmitt Chevrolet, Inc. v. Southeastern Fidelity Insurance Corp., 1992 WL 212008, 17 Fla.L. Weekly S579 (Fla. 1992), particularly when a term is not explicitly defined or clarified by the policy. Grissom v. Commercial Union Ins. Co., 610 So.2d 1299 (Fla. 1st DCA 1992).

Courts may not rewrite contracts or add meaning to create an ambiguity. State Farm Mut. Auto. Ins. Co. v. Pridgen, 498 So.2d 1245, 1248 (Fla. 1986). There must be "a genuine inconsistency, uncertainty, or ambiguity in meaning [that] remains after resort to the ordinary rules of construction." Excelsior, 369 So.2d at 942. Further, ambiguity is not invariably present when a contract requires interpretation, Weldon, 605 So.2d at 915; Gulf Tampa Drydock, 757 F.2d at 1175, and failing to define a term does not create ambiguity per se. Jefferson Ins. Co. v. Sea World, Inc., 586 So.2d 95, 97 (5th DCA 1991). Nevertheless, "[t]he insurer cannot, by failing to define the terms ... or to include any additional qualifying or exclusionary language, insist upon a narrow, restrictive interpretation of the coverage provided." National Merchandise Co. v. United States Auto. Assoc., 400 So.2d 526, 530 (Fla. 1st DCA 1981). Given these rules of construction, we turn now to the insurance policy before us.

B. Determination of Ambiguity

Ordinary rules of construction require us, first, to assess the natural or plain meaning of the policy language. E.g., Landress Auto Wrecking Co., Inc. v. United States Fidelity Guaranty Co., 696 F.2d 1290, 1292 (11th Cir. 1983). We must strive to give meaning to every provision, so we must look at the phrase "considered experimental" to determine its meaning within the policy. The policy does not define the term "experimental." The plain language of subparagraph (b), the provision relied upon by the district court to exclude coverage, does not indicate who will determine whether a proposed treatment is considered experimental. Further, it also omits standards for how that determination will be made. Thus, the contract phrase "considered experimental," without more, gives rise to a genuine uncertainty about who will determine whether a particular treatment is experimental and how that determination will be made. Indeed, even Mutual of Omaha concedes that there could be more than one source of such a determination. Ambiguity in the contract terms arises from the contract's silence on definitions.

The absence of a definition does not create ambiguity per se, as noted above. Nonetheless, the insurance company cannot then claim the narrow and favorable interpretation that its determination, alone, is controlling. In this case, Mutual of Omaha took upon itself the determination of whether the proposed treatment is considered experimental. While that is a reasonable interpretation of the policy language, it is not the only reasonable interpretation. It would also be reasonable to allow the determination of what is considered experimental to be made, for example, by the relevant medical specialists, according to generally accepted standards of community practice, or by a national association.

Mutual of Omaha's interpretation excludes coverage, but other reasonable interpretations may result in coverage. For example, although in its order in this case the district court did not assess whether HDC-ABMT is considered an experimental treatment by the relevant medical specialists, other courts have found that the relevant medical specialists do not consider HDC-ABMT experimental. E.g., Adams v. Blue Cross/Blue Shield, Inc., 757 F. Supp. 661, 663 (D.Md. 1991) (at the time Blue Cross denied coverage, Maryland oncologists generally acknowledged HDC-ABMT as accepted medical practice).

The existence of differing interpretations of the term "experimental" provides further evidence of a genuine ambiguity. Some courts have found that ambiguity is cured when the term "experimental" is qualified by specifying who will determine the experimental nature of the treatment or by specifying the standards by which the determination of experimental will be made. For example, in Fuja v. Benefit Trust Life Insurance Company, 809 F. Supp. 1333 (N.D.Ill. 1992), the insurance contract provided coverage for medically necessary treatment, which was defined as treatment that is, among other criteria, "not deemed to be experimental . . . by any appropriate technological assessment body established by any state or federal government." Id. at 1336. The court noted that a predecessor version of this criterion had been considered ambiguous, but that the qualifier "deemed ... by [a] technological [assessment] body" largely removed any ambiguity. Id. at 1340 n. 4. The court also noted that the National Cancer Institute, an appropriate technological assessment body under the contract, stated in its literature that "the term . . . `experimental' cannot be simply or unambiguously used to define new cancer therapies." Id. at 1340.

Similarly, in Boland v. King County Medical Blue Shield, 798 F. Supp. 638, 640 (W.D.Wash. 1992), a collective bargaining contract excluded experimental or investigational procedures, which were defined as:

any services so classified by the National Blue Cross and Blue Shield Association and any service whose use is still under clinical investigation by health professionals or is not generally recognized by the medical profession as tested and accepted medical practice. This exclusion also applies to items requiring Federal Drug Administration or other governmental agency approval if such approval was not granted at the time the service or supply was ordered.

The court determined that the contract unambiguously relied on a third-party classification of what was experimental or investigational, and, because the National Blue Cross and Blue Shield Association had classified the proposed treatment as investigational, granted summary judgment for the insurance company. Id. at 646.

Other courts have found that the term "experimental" is ambiguous when it is undefined in medical insurance contracts. For example, in holding that Blue Cross-Blue Shield's policy excluding experimental procedures was ambiguous, the United States District Court for the Eastern District of Virginia noted that the policy did not define "experimental" and did not explain what evidence would be used to demonstrate whether a particular treatment was experimental. Pirozzi v. Blue Cross-Blue Shield, 741 F. Supp. 586, 589-90 (E.D.Va. 1990). See also Nesseim v. Mail Handlers Benefit Plan, 792 F. Supp. 674 (D.S.D. 1992) (granting a preliminary injunction and requiring coverage for HDC-ABMT because the terms of the contract were ambiguous). Indeed, in Johnson v. District 2 Marine Engineers Beneficial Association, 857 F.2d 514, 516 (9th Cir. 1988), the court stated that "[i]n the context of modern medicine, the term `experimental' seems clearly ambiguous on its face." We agree that, in the context of a major medical insurance policy, the term "experimental," and in this case the phrase "considered experimental," is ambiguous when it is undefined.

IV. CONCLUSION

We need not reach the factual question of whether HDC-ABMT treatment for breast cancer is considered experimental. We hold that the district court erred by concluding that the term "experimental" is not ambiguous. The insurance policy does not clearly specify who will determine whether a treatment is considered experimental or how that determination will be made.

Accordingly, we vacate the district court's order denying Dahl-Eimers' application for a preliminary injunction and remand to the district court to determine whether there is a substantial likelihood that she will prevail on the merits.

VACATED and REMANDED.