Ion channels come in two main flavours, voltage-gated and ligand-gated, although voltage-gated ion channels can have binding sites for modulatory compounds, many nasty plant toxins, and an awful lot of the toxins produced by invertebrates, and cnidarians affect voltage gated ion channels.
Examples of such include the famous tetrodotoxin, puffer fish poison, of fugu fame, that imfamous sushi produced from very, very carefully prepared blowfish, which blocks TTX-sensitive sodium channels (and also found in the venom of the blue-ringed octopus, the california newt, a certain species of marine flatworm, and toads of the genus Atelopus)
TTX blocks those sodium channels, preventing an action potential (the current passed though nerves when the ion channels open and the cell fires in response to stimulation), causing paralysis, saxitoxin, or paralytic shellfish poison acts in the same way, although at a different binding site, whereas veratrine and related alkaloid toxins from Hellebore species, along with the venoms of the deadly, and very bad-tempered brazillian wandering spider, along with delta-atraxotoxin and the versutoxins of the funnel webs spiders in of the genus Atrax and Hadronyche lock sodium channels open, causing spontaneous, repetitive firing of nerves, resulting in a massive adrenergic storm, causing muscle cramps, sweating, severe tachycardia and severe increase in blood pressure, along with possible pulmonary oedema. Most scorpion toxins act by fucking with voltage-gated sodium or potassium channels (with the exception of the species Hemiscorpius lepturus, which is exclusively cytotoxic, and rather than neurological symptoms, causes horrendous flesh-eating effects, like that caused by recluse spider bites)
There are two main superfamilies of receptor expressed on cell surfaces (I am ignoring nuclear receptors located within cells here), metabotropic receptors such as GABAb do not form an ion channel, and instead of directly inducing ion currents within cells, signal more slowly than ionotropic receptors, and mediate their effects by inducing the release of second-messenger chemicals within the target cell, such as enzymes, transcription factors etc.
Ionotropic receptors on the other hand, such as GABAa, and the NMDA, AMPA and kainate type glutamate receptors signal far faster, and like other receptors are composed of complex protein subunits, arranged in such a manner as to form a central pore. this is the ion channel, and activation of the receptor by binding of an agonist drug/the natural neurotransmitter to those protein subunits, which can be complex enough to have multiple different binding sites, always one for the endogenous neurotransmitter, but often one or more allosteric modulatory sites causes a change in the 3d conformation of the receptor protein/ion channel complex which allows the channel to open for a time, and pass an ion flux.
The ionotropic receptors are those which express a ligand-gated channel, and induce the firing of an action potential by the binding of an agonist, voltage-gated ion channels do not depend on binding of a neurotransmitter or drug, but instead, the chance of firing an action potential depends on the voltage potential the cell is at, after a certain threshold is reached through sufficient depolarization of the cell, the voltage-gated ion channel opens, and passes current.
An allosteric binding site is a recognition site seperate from that for the main neurotransmitter, and can bind either positive (stimulatory in nature of the receptor expressing it) or negative (doing the reverse, inhibiting neurotransmission) and is dependent upon binding of a ligand at the orthosteric site, the main binding site for neurotransmitter.
GABAa is a good representative receptor of the ligand-gated ion channel flavour, it is made up of 5 seperate subunits, which form a central pore, and which passes chloride ions. Different subunit pentameric combinations result in receptors selective for different allosteric ligands, of which there are many types, and which expresses a large variety of allosteric sites, including that which binds benzos, another for barbs, one binding loreclezole and one of the active compounds from the sedative herb valerian, one for neurosteroids such as pregnanolone, and negative modulator recognition sites also.
These are seperate from the binding site that binds the endogenous neurotransmitter GABA, the main CNS inhibitory neurotransmitter (damping down other signalling processes and neurotransmitter release), an allosteric modulator of any kind will not cause any action in and of itself, if applied to a cell in the absence of the orthosteric ligand no effect will ensue, benzos work by potentiating the action of bound GABA, increasing the frequency of channel opening, whereas barbiturates more directly gate the chloride channel, and hold it open for longer, allowing longer duration of the resultant ion flux, and are thus more dangerous, as the body has far less ability to compensate for overactivity at the receptor.
Applied on their own though, in an environment free of GABA, benzos, barbs etc will do sod all in the way of causing current flow, not unless GABA is also bound to the main site, or another agonist.
An example of a couple of representative orthosteric agonists at GABAa (there are not nearly as many direct agonists of this type as there are positive allosteric modulators, the structural requirements for an orthosteric GABAa agonist are much more demanding) include muscimol, the active psychedelic-ish, dissociative-ish drug from the fly agaric mushroom, Amanita muscaria, isoguvacine, and isonipecotic acid.
Sometimes an ionotropic receptor can have some characteristics of both ligand, and voltage-gated activity, the NMDA-type glutamate receptor targeted by the dissociative anaesthetics such as PCP, ket and memantine is such a beastie. Activation of the NMDAr requires binding of either glutamate or aspartate, the endogenous neurotransmitters, and the main excitatory neurotransmitter expressed in the brain (along with a co-agonist, glycine, at an allosteric binding site, and there is also another allosteric site recognising polyamines), once the agonists bind, activation of the receptor causes a magnesium ion, which at the resting electrical potential of the cell expressing the NMDAr sits in the ion channel, blocking current flow to be spat out, in a voltage-dependent manner allowing the now open channel to pass current.
There are a great many seperate kinds of ion channel of the voltage-gated flavour, the main superfamilies being channels for sodium, calcium and potassium, and within those families, there are a crapload of subtypes, with varying sensitivities to different ion channel toxins, for example, I will use the calcium channel.
A non-exclusive list, includes N-type calcium channels, which are expressed in neurons, along with P and R types also found within the brain. L-type channels which are involved in mediating muscle contraction, such as that of the heart (medicinally important drugs acting here are calcium channel blockers, heart medications which act to slow the heart rate down by inhibiting opening of L-type Ca++ channels), T-type and a couple of others.
Some of these are sensitive to say, tetrodotoxin, saxitoxin etc, whilst being insensitive to other drugs/toxins, for instance the neuronal N-type channel is powerfully blocked by omega-conotoxin derived from the venom of the venomous marine cone snails, whilst the L-type channels are highly resistant to blockade from omega-conotoxin, but are blocked by the heart drug nifedipine, whilst nifedipine doesn't block the neuronal type recognising ω-conotoxin.
Does that clear things up some? Valproate blocking Na+ channels (sodium channel activation is excitatory in nature, I don't know if this is exclusively the case, but it is certainly vastly the predominant nature) will decrease neuronal excitability, and result in the requirement of a greater stimulation of the nerve before an action potential can be fired, thus having an overall stabilizing effect on neurotransmission, and acting to prevent or decrease the propagation of the uncontrolled shitstorm of random nervous firing that is characteristic of an epileptic seizure.
Incidentally, valproate was discovered by accident, it used to be thought inert and be used in neurochemistry experiments on animals (uggh!) merely as a delivery solvent for drugs being tested, it was only discovered when eventually it was used to deliver a convulsant poison to an animal, and it was found to have lesser effect than expected, exposing it as an antiseizure drug.
Don't know if you are female or male, but do not take during pregnancy, it is well known to be teratogenic, and cause lots of nasty mutations (bad thing) and increase the frequency of autism (good thing)