Search Console shows you forty queries. Nineteen of them are the same three phrases in different spellings. Greek keyword research is mostly the work of seeing that, then merging: one page per intent, with every spelling of that intent pointing at the same URL.
What Google actually matches
Google generally treats accented and unaccented Greek as one query. Search for «ενοικίαση αυτοκινήτου» and «ενοικιαση αυτοκινητου» and you will usually get the same results page. The tonos is not a ranking factor. The final sigma behaves the same way: «σελιδας» and «σελίδας» are one word to the index.
That does not make spelling irrelevant. It matters in three places.
Your analytics. Search Console and GA4 report exactly what a person typed. If you only track the accented form, the unaccented queries look like a separate, smaller keyword and you will under-invest in the one that actually brings traffic.
Your ads. Google Ads matches close variants, but an unaccented or Greeklish query is a different string, and the match is not always as generous as people assume. Negative keywords and exact-match campaigns especially reward you for listing every spelling you can find.
Your copy. The title a human reads should carry correct accents. «Ενοικιαση αυτοκινητου» in an H1 reads like a typo, and a typo in the first line costs more trust than a keyword gains.
Test your own phrases instead of trusting a rule of thumb. Google’s matching changes, and brand names behave differently from generic phrases. The Search Console documentation is the place to check what a page is actually indexed for, and Google’s overview of how Search works explains why the query and the page text are matched, not compared letter by letter.
Greeklish: two systems, one audience
Greeklish is Greek typed with Latin letters. A meaningful share of Greek searches still arrives that way: older phones, a keyboard set to English, someone in a hurry, Greeks living abroad.
There are two systems, and they produce different strings.
ELOT 743 is the official transliteration. Vowel digraphs map predictably: αι→ai, ει→ei, οι→oi, ου→ou — and the consonants have fixed values: θ→th, χ→ch, ψ→ps, ξ→x, μπ→mp, ντ→nt, γκ→gk, τσ→ts, τζ→tz. Two rules surprise people: αυ and ευ become af/ef before a voiceless consonant, and υ is written y.
Phonetic Greeklish is what people actually type. η, ι and υ all become i; ω becomes o; αι becomes e; μπ becomes b; ντ becomes d; γκ becomes g.
Here is the same phrase in all four forms, which is exactly what our Greek keyword variants tool produces:
| Spelling | Result |
|---|---|
| With accents | ενοικίαση αυτοκινήτου |
| Without accents | ενοικιαση αυτοκινητου |
| ELOT 743 | enoikiasi aftokinitou |
| Phonetic | enoikiasi autokinitou |
The ELOT form is not wrong. It is a standard nobody types. «αυ» before a voiceless «τ» becomes «af», and «υ» becomes «y», which is how you end up with «aftokinitou». That is why you keep the official form for slugs and file names, and the phonetic form for campaigns and anything a person will read.
The workflow we use
- Export the queries you already have. In Search Console, open Performance → Queries for the last 16 months and export. This is the only list that comes with evidence.
- Expand it in Keyword Planner. Take the phrases that already have impressions and let Google suggest the neighbours. Do not start from a blank page; start from queries that already found you.
- Run each phrase through the variants tool. Copy the accented, unaccented, ELOT and phonetic forms into a spreadsheet.
- Merge by intent. One row per intent, with every spelling in separate columns. If two rows mean the same thing to a customer, they are one row.
- Map one intent to one URL. The page that answers the query gets the intent, plus its variants in the title, the H1 and once in the body, written naturally rather than as a list.
The merge step is where most Greek sites lose. Four spellings spread across four pages produce four thin pages that compete with each other. One page with four spellings produces one page that ranks.
Intent traps that cost Greek sites traffic
Singular versus plural. «ενοικίαση αυτοκινήτου» is someone looking for one car now. «ενοικιάσεις αυτοκινήτων» can be a business directory query or someone comparing companies. They deserve different pages, or one page with a section that answers both, not a keyword-stuffed paragraph.
«τιμή» versus «κόστος». Price is what a customer pays; cost is what it takes to build. A page about «κόστος ιστοσελίδας» that only lists packages answers half the query. Say what drives the number up and down.
Brand plus city. «JupiWeb Ηράκλειο» is a navigational query. It belongs on the contact page or the home page, not on a new landing page created for every town.
Transactional queries on informational pages. A hotel or car-rental search belongs on a landing page with availability, dates and a booking path. If the best page for that query is a blog post, the site is asking the reader to hunt for the button.
What this looks like on a project
When we built the site for Kourites Car Rental in Stalis, Crete, the brief was not “rank for car rental”. It was one page per rental intent — car rental in Crete, rental at the airport, rental for a specific season — with the Greek and English spellings of each intent handled on the same page.
The same logic applies to budgets. A business website starts at €490 and takes about 14 working days; an e-shop starts at €890 and about 25 working days; a second language adds €129. Prices exclude VAT, and they are fixed, not hourly. Keyword work sits inside that scope: we do not sell a separate “SEO package” before the pages exist.
What to read next
If you are still deciding what SEO is meant to do for a small business, start with SEO explained for small businesses. If your pages already exist and simply do not appear, read why your page is not on Google. It is usually an indexing problem, not a keyword problem.