The Quick Reference You Actually Need
Most people think two letter words in the english language are just an odd trivia category, but they show up constantly when you're working with constrained systems. Username generators, password policies, crossword puzzles, SMS character limits, and even some legacy database fields still force you to deal with them. I spent weeks debugging a script that randomly generated valid abbreviations for an old messaging system, and the word list itself was more of a headache than the code. The official Scrabble dictionary, NASPA Word List, and Collins Scrabble Words all agree on roughly 110 to 115 valid two-letter words depending on which list you use. The core set has barely changed in decades. Here is the working list most people end up relying on: A, Ah, Al, Am, An, Ap, Ar, As, At, Ay, Ba, Be, Bi, Bo, By, Da, De, Di, Do, Eh, El, Em, En, Er, Es, Et, Ev, Ex, Fa, Fe, Go, Ha, He, Hi, Ho, Hu, I, If, In, Is, It, Ja, Ka, Ki, Lo, Ma, Me, Mi, Ml, Mo, Mu, My, Na, Ne, No, Nu, Oa, Ob, Od, Of, Oh, Oi, Ok, Ol, Om, On, Op, Or, Os, Ow, Ox, Pa, Pe, Pi, Po, Qa, Qi, Re, Sh, Si, So, Ta, Te, Ti, To, Uh, Uo, Us, Ut, We, Wo, Xa, Xi, Ya, Ye, Yo, Za.
If you are building something that validates input against this set, do not try to hard-code it by hand. I learned that the hard way when a typo in "Qi" cost me three hours of tracing a false negative bug. Download a verified word list from a reliable source or use a published word game dictionary file instead of recreating it yourself.
How People Actually Use These Lists
The most common practical use is validation logic. A simple inclusion check against a stored array or hash set is all you need. In practice I usually load the list as a JSON or plain text file and run a set lookup, which takes under a millisecond even at scale. Another frequent use case is generating short codes, passwords, or identifiers where exactly two characters matter. In those cases you want the full valid set because common trigrams like "th" and "he" dominate regular language but do not necessarily show up as standalone two-letter words. Scrabble players use a slightly different subset depending on the edition. NASPA keeps a North American tournament list, while Collins covers UK and international play. The difference between the two matters if you are building a multi-regional word game backend. I ran into this when a player reported that "Zh" was valid in one match but rejected in another. The fix was just normalizing which dictionary file the app loaded per region.
Get the Full Details
Edge Cases That Will Bite You
The biggest problem area is proper nouns and abbreviations that slip into casual play. Words like "Ma," "Pa," "Pi," and "Ti" are legitimate Scrabble entries but they confuse people who assume they are abbreviations. Similarly, "Qa" and "Qi" look like typos until you check the list. I once built a form validator that accidentally excluded "Ai" because an intern thought it was a greeting and filtered it out. It took me a while to trace the issue since "AI" in uppercase is obviously a common acronym, but the lowercase entry "ai" is actually a valid word in the list. Another tricky area is loan words and archaic terms. "Es," "ex," "fa," "ha," "sh," and "si" feel informal or non-standard to many native speakers, yet they are all accepted in standard dictionaries. If your application validates words for educational purposes, you should be upfront about which dictionary the tool follows so users do not get confused when a word they consider slang passes validation.
Implementation Notes
When implementing a lookup, use a set data structure rather than a list. Linear search over 110 items is fast enough on a single request, but if you are validating thousands of entries in a batch, the difference becomes noticeable. A hash set gives you O(1) lookups and keeps your code clean. Store the list in a separate configuration file so you can swap dictionaries without touching the core logic. If you are writing this in JavaScript, a simple approach is to keep a constant array and convert it to a Set at module load time. In Python, a frozenset works well for the same purpose and uses less memory than a standard set if the data is never mutated. For backend services, caching the validation result per word is usually pointless because each lookup is already sub-millisecond. The real bottleneck tends to be I/O when loading large files, not the comparison itself. I have also seen teams use regular expressions like ^[aeiou...etc]$ with every two-letter word, which works but makes the regex unwieldy and harder to maintain. A plain set is easier to audit and far less error prone. If you must use regex for some reason, generate it programmatically from the word list instead of writing it by hand.
Common Pitfalls
One frequent mistake is assuming that every two-character string is valid just because it appears in a word. "Th" is a common digraph but it is not itself a word. "Er" is a word, but "re" and "er" are separate entries with different meanings and uses. Confusing substrings with full words leads to inflated validation scores and incorrect answers in games. A second issue is case sensitivity. Some systems treat "I" and "i" differently, but the standard lists include "I" as a standalone word and treat lowercase variants case-insensitively in most word game contexts. Normalize input to lowercase before lookup to avoid unnecessary rejections. I once had a client complain that their word checker kept rejecting "A" as invalid because the input came in lowercase from a mobile keyboard and their code did not normalize it. A third problem is outdated dictionary versions. The official word lists get updated occasionally, and new entries appear while others fall out of favor. If your tool claims to use the NASPA list but ships with a five-year-old file, players will catch the discrepancy. Always note the dictionary version and date in your documentation.
Where This Approach Falls Short
A static word list only covers exact matches. It does not help with partial matches, fuzzy validation, or handling hyphenated terms. If you need support for dialect-specific words beyond the major Scrabble dictionaries, you will need to supplement with additional sources. Regional variations in British and American English mean that some two-letter entries accepted in one market may be rejected in another, and there is no single universal standard that satisfies everyone. For serious projects, I recommend pairing the word list with a versioned configuration and a clear citation of the source dictionary. That way when a dispute comes up, you can point to a specific revision instead of arguing over which list someone remembered from childhood. I keep a changelog for my own projects, and it has saved me from multiple unnecessary escalations with users who swear a certain word should be valid.
