SEO tools: which one for which job
Choosing an SEO tool means first deciding which question you want answered, then going to the tool that answers that one question. There are six main categories and each covers exactly one kind of question; the rest of the differences between tools in the same category are taste and price.
- Lesson 13 of 15
- Intermediate
- Free, no signup
Six tool categories, around the job you want done
Each category answers one kind of question. Tools inside a category compete; tools from two different categories do not replace each other.
No category substitutes for another, and none of them says why. The why comes from the page and the customer.
Last checked: Facts and tool names in this lesson are re-checked against their sources on this date.
Why do most people buy the wrong tool?
Because they run the sequence backwards: first they subscribe to a big platform, then they look for something to do with it. The result is a dashboard with forty reports that gets opened once a month and produces no decision at all.
The right sequence is simple. Have one specific question: "which of my pages has the wrong canonical", "why is this page not being indexed", "where did my competitor get that link", "what phrase do people search for this service". Each of those belongs to one category of tool, and most of them are answered by a ten-second check rather than a monthly subscription.
One point that gets said less often: tools split into two families, and that split matters more than the brand names. The first family measures something visible right here and right now: the response headers of a URL, an SSL certificate, a DNS record, the tags on a page. The second family estimates something stored in somebody else's index: search volume, domain authority, a competitor's link list. The first is exact and costs nothing to give away. The second is always an estimate and always costs money, because somebody had to build that index.
Know that one distinction and you stop expecting competitor data from a free tool, and you stop expecting an expensive tool to tell you what your page's canonical is.
Six categories, and the question each one answers
Each category has one question and one thing it does not know. That second half is usually missing from the product page.
Rank tracking. The question: "where am I in the results for this phrase right now?" It does not know why you are there, and the rank it reports depends on city and device. For your own real click data the right place is Search Console rather than a tracker.
Keyword research. The question: "what do people type for this subject, and how much?" It does not know the intent of the person typing it, and its number is an estimate rather than a count. The keyword research lesson is where this goes deeper.
Technical and page. The question: "what is this URL telling Google?" Headers, redirects, the canonical, meta tags, structured data, robots and the sitemap. This category estimates nothing; it reads exactly what your server returns right now. Which is precisely why it can be free.
Links. The question: "who linked to me or to my competitor?" No tool has the complete list, because none of them has Google's index; each has its own, and that is why their numbers disagree. The link building lesson opens that up.
Competitors. The question: "which keywords does my competitor rank for that I do not?" This category is the most expensive because it needs the largest index, and in the Iranian market it is also the least accurate; the reason comes in the section on numbers.
AI and content. The question: "how do I produce and check this text or table faster?" This is the newest category and the most oversold. Its real place is mechanical work: sorting, clustering, converting formats. Where it fails is anywhere it has to know something and does not.
Our own toolbox holds 24 tools arranged in six groups with slightly different lines: keywords, competitors, links, domain, page and rank. If you want to see each one up close, the SEO tools list is right there.
Is the data you want about your own site, or about somewhere else?
- Headers, redirects, status codes
- Canonical, meta tags, schema, hreflang
- SSL certificate, DNS records, domain record
- Real clicks and impressions, from Search Console
Estimated from someone else's index, paid
- Search volume and cost per click
- A competitor's ranking keywords
- The list of linking domains
- An authority score, which is that company's metric
This question nearly always settles whether you have to pay, and much sooner than comparing prices does.
Which checks can you run for free right now?
There are eight checks that run on our own site with no signup and no payment, and the reason is not generosity; it is structural. Those eight make no call to an outside data service and run on our own server, so each run costs one network request rather than money. The other 16 tools are wired to a paid data provider and every run genuinely costs something. That is the whole rule.
And notice that all eight sit in the "measuring" family described above: all eight answer questions about your own site, not about a competitor.
| Check | The question it answers |
|---|---|
| Meta tags and canonical | What are this page's title, description and canonical |
| Redirects and headers | Where does this URL go, and with which status code |
| robots.txt and sitemap | What is blocked, and is the sitemap healthy |
| Structured data | What schema does this page actually print |
| hreflang | Do the language versions point at each other correctly |
| SSL certificate | Is the certificate valid and when does it expire |
| DNS records | Where does the domain point and is its mail healthy |
| Domain record | When was the domain registered and when does it expire |
A practical rule falls out of that table: before hunting for a paid tool for any problem, run those eight checks on the page in question first. Our experience is that a large share of what gets reported as "an SEO problem" is answered by one of those eight lines: a wrong canonical, a redirect chain, a leftover noindex, a sitemap still listing old URLs.
How real are the numbers these tools give you?
There are three kinds of number in these tools and they are not the same. A measured number, like a URL's status code or a certificate's expiry date: you can check it and that is that. A number reported by a source, like clicks and impressions in Search Console: it comes from Google itself but carries its own rules. And an estimated number, like search volume or domain authority: a model built on a private company's index.
That third kind is where people go wrong most. Domain authority, or any score like it, is not a Google metric; it is that company's metric. Two tools give two different numbers for one domain and neither is wrong, because they are measuring two different things.
And now the part specific to the Iranian market that no product page mentions. For a Persian phrase, the data provider we use has no country filter at all; the number you get is the worldwide search count for that Persian phrase. We measured eight phrases and the share of Afghanistan, the only other Persian-speaking country, came out at about one percent, so that number is effectively Iran's number. The reverse is more interesting: when a country filter is applied the number roughly halves, because half of Iranian users' searches are recorded from outside Iran. So the unfiltered number is closer to reality than the filtered one.
And one complete absence: keyword difficulty and search intent are not available for a Persian phrase. Our keyword trend tables leave those two columns out of the Persian rows for exactly that reason. If a tool hands you a difficulty score for a Persian phrase, ask which database that number came from.
The practical consequence: for a Persian site the number you can lean on is the one you measure yourself, plus what Search Console reports from your own real traffic. The rest show direction rather than magnitude, and deciding on direction is fine; reporting them as an exact magnitude is not.
Which number you can lean on
Reproducible, so you can lean on it
- The status code and target of a redirect
- The canonical the page prints
- A certificate expiry date and a DNS record
- Your own site's clicks and impressions in Search Console
Estimated, so direction only
- The monthly search volume of a phrase
- A domain authority score, from any tool
- The total link count of a site
- Keyword difficulty, which does not exist for Persian at all
The second column is not worthless; it shows direction. The trouble starts when it gets reported as an exact magnitude.
Where to start, and where to spend money
The order we follow on real projects, and it is enough nine times out of ten: connect Search Console first, because it is the only source of your real click data and it is free. Then run those eight checks on your most important pages and fix whatever turns up. Only after those two steps go to a paid tool, and only for one specific question the first two steps did not answer.
What should you pay for? Data you cannot produce yourself: a competitor's ranking keywords, a list of linking domains, the search volume of a market. Refusing to pay for something a single HTTP request answers is not thrift; it is not knowing the difference.
And a position not everybody shares: for a small Iranian business, a monthly subscription to a large foreign platform is usually premature. Not because of the price, but because its data carries the same limits for the Persian market described above, and because you have not yet collected enough data from your own site to decide anything with. Collect six months of Search Console first. Then look at which question is left that you cannot answer, and let that one question tell you which tool to buy.
The fast path, with AI
The fast path here is not asking a model which tool to buy. It is taking four base checks in a few seconds from the terminal itself, then handing the raw output to a model so it builds a prioritised list where every item is pinned to a line of that output. The eight free checks above do exactly this with a graphical interface, and if the terminal is not your thing, use them instead; but if you have ten URLs to go through, this is faster. A cheap fast model is enough; our current pick is in the <a class="text-link" href="/en/ai/">AI section</a>.
- Run the first command on the domain you want to check. It gathers four things: the redirect chain and final code, a few important headers, the SSL certificate, and the DNS records plus the top of robots.txt.
- Copy all of the output verbatim and hand it to the model with the stage two prompt. Summarise nothing and edit nothing; the entire point is that the model works on the real text rather than on your account of it.
- Judge the output by its own rule: any item without an evidence line is not an item and gets deleted. That is the only filter you need.
- For every item that survives, open the page yourself and look. The terminal output says what is there; whether that thing is a problem on this site is your call.
Copy-ready recipe
D="example.com"; U="https://$D/"
echo "== redirects and final code"
curl -sIL -o /dev/null -w '%{http_code} %{url_effective} redirects=%{num_redirects}\n' "$U"
echo "== headers"
curl -s -o /dev/null -D - "$U" | grep -iE '^(server|content-type|cache-control|x-robots-tag|location):'
echo "== SSL certificate"
echo | openssl s_client -connect "$D:443" -servername "$D" 2>/dev/null \
| openssl x509 -noout -subject -issuer -dates
echo "== DNS and robots"
dig +short "$D" A; dig +short "$D" MX; dig +short "$D" TXT | grep -i spf
curl -s "https://$D/robots.txt" | head -20
---- Stage 2: the model prompt ----
Role: technical SEO reviewer. You have only the output below and know nothing
else about this site.
Raw output:
{everything the command above printed}
Write a prioritised list, at most eight items, from most to least serious.
Each item must carry exactly three things:
- the finding, in one sentence
- the evidence line, copied verbatim from the output above
- what should be done, in one sentence
Rules:
- Write no item whose evidence line you cannot take from the text above. Delete
that item entirely.
- Assume nothing that is not in the output. If the page tags are not here, for
instance, write nothing about the title and description.
- Write no generic SEO advice. Only what is visible in this text.
- If the output is healthy, say so and return an empty list.Before you trust the output: Keep three things for yourself. First, these four checks only see the outer layer of the site: in-page tags, schema and hreflang are not in them, and for those you go to the page checks. Second, a healthy output does not mean a healthy site; it means these four things are fine. Third and most important: any item the model writes whose evidence you cannot find in the text was invented, however reasonable it sounds. Enforce the "an item without evidence is deleted" rule yourself rather than leaving it to the model.
AI in this kind of work
Among SEO tools, AI is not a separate category; it is a layer that sits on top of everything else's output. What it genuinely does well is read raw output and organise it. What it does badly, and what everyone asks of it, is to decide for you which tool to buy, or to produce a number it does not have.
Tools that actually help
- Claude Suits the "every item needs an evidence line" stage, because it paraphrases less when you ask for verbatim quotes. Iran is on neither of Anthropic's two supported-countries lists.
- Gemini Easier with long, multi-domain output. Our own entry notes Iran is not on Google's supported-countries list; the payment route is in <a class="text-link" href="/en/ai/buy/">buying AI access</a>.
- RGB SEO tools Not AI, and deliberately in this list: every AI layer works on somebody else's output, and these eight checks hand you that output with no terminal and no signup.
Where it backfires
The specific risk in this subject is invented features and invented numbers. Ask a model which tool does a certain thing and it will build you a list in which some rows do not exist or do not have that capability; because the question looks like questions that have answers, the answer comes out looking like an answer. The worst case is asking a model for a number: domain authority, search volume, a link count. The model is connected to none of those indexes and the number it gives came from nowhere. For a Persian phrase the risk is higher still, because as described above keyword difficulty is not available for Persian at all, and the model does not know that and produces a number. The second risk is about data: the raw output of a technical check is your client's site structure. For a public site it does not matter, and for a project that has not launched yet, sending it to an outside service is a decision; whether the conversation is used for training depends on that service's plan and settings and is written on its own page.
Sources: Anthropic: is my data used for model training Google: Gemini Apps Privacy Hub Search Console Help: Performance report, dimensions and data groupings
Where this advice stops
No tool tells you how much money your business made from that traffic, and no tool tells you why a ranking changed. Measuring tools only see the moment they were run in, and estimating tools are exactly as accurate as the index behind them. For the Persian market those indexes are thinner: keyword difficulty is unavailable and search volume shrinks misleadingly under a country filter. And none of this is solved by buying a more expensive subscription.
From our own work
The Persian numbers we publish on this site come from a measurement decision we reached ourselves, and it is written on the pages that carry them. We take the Iranian rows of our trend tables from Google Ads through DataForSEO, with no country filter. The reason is that the service has no country filter for Persian at all, so the number is the worldwide search count for that Persian phrase; we measured eight phrases and the share of Afghanistan, the only other Persian-speaking country, came out at about one percent. The other side matters more: apply a country filter and the number roughly halves, because half of Iranian users' searches are recorded from outside Iran. Which means a report built with an Iran filter shows the market at half size. And the keyword difficulty and search intent columns are not printed at all in our Persian tables, because they are not available for a Persian phrase; the three non-Persian rows in those same tables come from the Turkish and US databases, where both columns do exist. All four of these are stated on the keyword trends pages and you can go and look at them now.
Real follow-up questions
Are free tools enough to start with?
For everything about your own site, yes. Search Console gives you your real click data for free, and the technical checks run for free too. What does not come free is competitor and market data, because somebody has to have built and maintained that index.
Why do two tools give one domain two different authority scores?
Because neither is a Google metric. Each company builds its own link index and defines a score on top of it, so the two numbers are two different things and comparing them to each other is meaningless. The meaningful comparison is between two domains inside the same tool.
Is the search volume a tool gives for a Persian phrase reliable?
As a direction yes, as an exact number no. First check whether that number was taken with a country filter, because an Iran filter roughly halves it and shows the market as smaller than it is. And if the same tool also hands you a difficulty score for a Persian phrase, that is where you should get suspicious.