Writing formulaic Website positioning content material, updating outdated articles, reporting on weblog efficiency, even operating difficult efficiency analyses… these are all issues Agent A does for me.
Listed below are a few of our favourite Agent A use instances for content material entrepreneurs.
Agent A is a advertising and marketing agent from Ahrefs—primarily an AI assistant with direct, unrestricted entry to the total Ahrefs dataset that may perform advertising and marketing and Website positioning duties autonomously, fairly than simply reply questions.
It’s a workspace the place a long-running AI agent builds, runs, and maintains instruments for you, on infrastructure it controls. It’s not only a chatbot you ask inquiries to. It’s nearer to a junior engineer who occurs to have a deep Website positioning background, a Postgres database, a Flask server, and entry to all of the frontier LLMs.
Right here’s how Agent A is completely different from ChatGPT, Claude Code, and even utilizing the Ahrefs API or MCP:
- Unrestricted entry to Ahrefs endpoints. Each endpoint we use to construct Ahrefs is on the market to Agent A, together with many you may’t entry through API or MCP: Key phrases Explorer, Website Explorer (101 endpoints alone), Model Radar, Net Analytics, GSC integration, AI Content material Helper, Website Audit, Rank Tracker, Content material Explorer, Batch Evaluation… You’ll be able to even create and replace tasks in Ahrefs simply by speaking with Agent A.
- Severe tech stack beneath. Agent A is sort of a advertising and marketing Swiss military knife, with all of the tech and devices you want to construct functions and studies within the proper manner. To not get too technical, however Agent A is aware of exactly how you can flip your huge concepts into Postgres databases, Flask apps, and scheduled cron jobs. It makes use of webhooks, an OpenRouter LLM proxy with 300+ fashions, internet fetch with full-page parsing, PDF extraction, OCR…
- Constructed-in connections to all of your advertising and marketing instruments. You should utilize native connectors for Slack, HubSpot, GitHub, Notion, Linear, Mailchimp, Resend, SendGrid, Stripe, Gong, WordPress, Airtable, Reddit, Apify, and even Semrush—in case you need to migrate to Ahrefs 😉
- Professional talent and app library. The Ahrefs workforce (myself included) has contributed pre-built advertising and marketing expertise and functions that allow you to automate essential advertising and marketing processes the exact similar manner an skilled Ahrefs energy consumer would.
Right here’s how we’re utilizing Agent A to automate content material advertising and marketing:
I used Agent A to construct the Weblog Pipeline, an 11-stage assisted-writing workflow that automates Website positioning content material creation. You enter a goal key phrase (or higher but, ask Agent A to search out one with the built-in ‘Content material Hole Evaluation‘ talent), and Agent A works sequentially by 11 separate phases to ship a publish-ready article draft.
Agent A works in sequence by key phrase analysis, SERP evaluation, AI Content material Helper matter snapshot, bulleted define, product mentions, drafting, inner linking and quotation sourcing, picture technology, and formatting for publication. You’ll be able to see the output at each stage and edit it in-line.
I hold tinkering with the Weblog Pipeline. My newest additions embody:
- Vibe modifying mode: now you can use a chatbox to provide the LLM directional suggestions on article outlines and drafts, and it’ll motion your modifications for you. No extra guide modifying or copy-pasting. All vibes, all of the time!
- Customized model guides: add your individual model information and choose an writer profile per article, skilled by yourself writing samples.
- Branded movement diagrams: counsel and generate movement diagrams in your model styling.
Starter immediate:
Construct me an assisted long-form article pipeline. Atomic enter is a goal key phrase. Levels run sequentially as background jobs the UI polls: (1) key phrase analysis through Ahrefs, (2) competitor SERP fetch, (3) AI Content material Helper matter snapshot, (4) bulleted define with mandated matter protection, (5) data-mention placement, (6) full draft, (7) polish, (8) WordPress shortcode formatting + .docx export. Every stage reveals its output, has an "edit" textarea, and a "refine with suggestions" chat that re-runs the stage with my notes. Fashion information comes from a per-author voice profile.
We have now over 1,000 articles on the Ahrefs weblog (together with localized variations), and maintaining all of them updated is greater than a full-time job.
I used Agent A to construct an automatic updating course of:
Share a broadcast URL, and the pipeline fetches the article, extracts the web page content material, then runs 4 diagnostic phases in parallel:
- Steerage: you set the scope of the updating course of—mild refresh vs. full rewrite.
- Claims audit: the LLM flags each statistic, examine reference, and dated assertion within the put up, grades every for staleness, and the place wanted, finds a alternative URL to reference.
- Ahrefs mentions: cross-checks the article in opposition to Ahrefs options launched since publication and suggests the place to drop new ones.
- Matter gaps: re-runs the SERP in opposition to present top-ranking pages and surfaces matters these pages cowl that yours don’t.
My favourite function is the Preview stage: a side-by-side diff between your present article and the proposed updates, with the choice to just accept/reject per change.
You’re by no means observing an AI-rewritten draft and attempting to identify what modified. You see the unique on the left, the proposed edit on the precise, and also you click on by them. The Replace Pipeline is what makes a month-to-month “refresh 20 outdated posts” dash truly manageable.
Starter immediate:
Construct me a blog-post replace pipeline. Enter: a broadcast URL. Fetch the article. Run 5 diagnostic phases: (1) Steerage — I set scope (mild refresh vs. full rewrite); (2) Claims Audit — LLM extracts each stat, examine reference, and dated assertion and grades every for staleness with a recommended alternative; (3) Ahrefs Mentions — cross-check in opposition to Ahrefs options launched since publication and counsel the place to drop new ones; (4) Matter Gaps — re-run the SERP, floor matters present top-ranking pages cowl that mine does not; (5) Authoritative Pages — discover linkable sources printed since my article. Ultimate stage: side-by-side diff between present article and proposed updates, with settle for/reject per change. Export the accepted model as markdown and WordPress shortcodes.
Each month I share a fairly detailed efficiency report for the Ahrefs weblog. It combines tons of knowledge sources, consists of a great deal of visualisations, and affords a layer of my (theoretically) professional evaluation.
The report used to take a full day to create. Now, Agent A generates it for me, routinely, on the 2nd day of every month (to permit time for GSC information to complete accumulating).
(That is simply dummy information—you may inform by the truth that it’s rising month-over-month…)
It pulls Google Search Console, Ahrefs Net Analytics, and the GSC dashboard chart collectively into one view with KPI tiles, 12-month pattern chart, subfolder splits, winners-and-losers tables, day by day anomaly callouts, and full paginated lists of each Website positioning and non-Website positioning put up.
I’ve additionally included an editable “month-to-month overview” subject. The AI doesn’t write the evaluation: that’s my job. As a substitute, it analyses the month’s efficiency and suggests 6-10 candidate bullet factors I can assessment and cherry-pick into the overview, in the event that they help my evaluation.
Starter immediate:
Construct me a month-to-month weblog efficiency report. Pull GSC + Ahrefs Net Analytics for the present month. Present KPI tiles, a 12-month pattern chart with a migration marker, subfolder break up, winners/losers tables (paginated, 25/web page), day by day anomaly callouts, and full paginated tables of each put up. On the prime, an editable markdown "month-to-month overview" with auto-save. Beside it, an AI panel that takes my cached KPIs + an "business context" textarea I fill with algo-update information and produces 6-10 candidate bullets I can copy. Add a "publish to public website" button that snapshots a read-only view.
I’ve at all times been fascinated about topical authority: the concept Google rewards web sites that cowl their space of experience in complete element. We have now an enormous, sprawling weblog, and I wished to see how our “off-topic” articles carried out relative to our core articles.
So Agent A ran the evaluation for me. To get a bit technical, the Weblog Semantic Audit mapped each URL on our weblog into vector house, computed a website centroid, and bucketed pages by their cosine distance into core/close to/mid/far.
Then it enriched each web page with Ahrefs site visitors information, so every bucket reveals you common natural site visitors, referring domains, UR, and key phrase protection. It even grouped associated articles collectively to focus on pure matter clusters.
The consequence solutions a query I at all times wished to reply: “Do my off-topic posts underperform my core posts, and by how a lot?” On our weblog, the reply is sure: core pages get roughly 2× the site visitors of far pages.
Starter immediate:
Run a semantic audit of my weblog. Pull each URL from the sitemap, fetch the content material, embed every web page (imply of passage embeddings) utilizing a 3072-d embedding mannequin. Compute the positioning centroid and bucket pages by cosine distance to it (core/close to/mid/far utilizing imply ± 1/2σ — not quartiles). Enrich every URL with Ahrefs batch evaluation (org_traffic, refdomains, UR, key phrases). Run k-means with silhouette scan (okay=2..12) to search out pure matter clusters. Output: bucket histogram, per-bucket Ahrefs averages, cluster summaries with pattern URLs, and a verdict on whether or not the weblog is tight or diffuse.
I spend lots of time reviewing competitor blogs and on the lookout for article inspiration. After I requested Agent A to assist, it constructed a software known as Competitor Feed.
Competitor Feed watches an inventory of competitor weblog sitemaps and surfaces new posts every day. Every put up is saved with a title, publish date, first-paragraph excerpt, and a one-line LLM abstract. I then triage every merchandise as save/dismiss/ignore: good concepts get saved into my backlog of content material concepts.
Whenever you save a put up, the app fires an Ahrefs Key phrases Explorer pipeline in opposition to the title: it extracts a 2-3 phrase seed matter, fetches key phrase solutions, ranks them by quantity and intent, and attaches the outcomes to the saved row.
So “competitor watching” stops being a passive feed and turns into an lively key phrase pipeline: each attention-grabbing competitor put up produces an inventory of key phrases we may go after on the identical matter.
Starter immediate:
Construct me a competitor weblog watcher. I configure an inventory of competitor weblog sitemap URLs. A day by day job diffs every sitemap, fetches new URLs, and for every new put up reveals title, publish date, first-paragraph excerpt, and a one-line LLM abstract of the angle. Triage states: new / saved / dismissed. After I save a put up, run an Ahrefs Key phrases Explorer pipeline in opposition to the title: extract a 2-3 phrase seed matter, fetch key phrase solutions, rank by quantity and intent, connect outcomes to the saved row. The output is competitor-inspired key phrase lists, not a passive studying queue.
Louise constructed the LinkedIn Scrapbook, her personal private swipe file for content material inspiration. You put in a Chrome extension; when you browse LinkedIn, one click on on any put up saves it, with full textual content, writer, engagement metrics, and media, into your scrapbook.
Scrapbook turns into very attention-grabbing once you’ve saved some content material, care of all the additional instruments Louise added, like:
- Trending key phrases: this reveals key phrases rising in recognition throughout the posts you’ve saved, so you may spot themes your community is gravitating towards earlier than they hit the Website positioning press.
- Content material hole: compares matters in your saved posts in opposition to matters you’ve printed, with the intent of surfacing “belongings you’re consuming however haven’t written about”.
- Instance finder: once you’re drafting an article and wish a related instance, you should utilize semantic search throughout the scrapbook to return related content material.
- Ask your scraps: question your database of saved snippets (with questions like “which scraps point out AI Overviews?”).
Starter immediate:
Construct me a LinkedIn swipe-file app with a Chrome extension. The extension provides a "Save to Scrapbook" button to each LinkedIn put up; one click on captures put up textual content, writer, engagement metrics, and media URLs and POSTs to my Console app. The Console app shops posts in Postgres with full-text search. Construct three instruments on prime of the corpus: (1) Trending Key phrases — extract matter seeds from saved posts, floor rising matters over a rolling window; (2) Content material Hole — diff matters in saved posts in opposition to matters in my printed weblog posts, output what I am consuming however have not written about; (3) Instance Finder — semantic search over the scrapbook with deep hyperlinks again to LinkedIn. Add a generic web-clipper extension too for non-LinkedIn URLs.
7. Get scientific inner linking suggestions
Inside linking is a kind of Website positioning chores that “ought to” get accomplished each time we publish and nearly by no means does.
So I obtained Agent A to construct the Inside Linker. Feed it a brand new article (both a broadcast hyperlink, or pasted draft markdown for unpublished items) and it finds essentially the most related present posts that ought to hyperlink to it.
Beneath the hood, it embeds the enter article with Gemini and cosine-compares in opposition to all different articles from our sitemap. The highest candidates then get rescored with a particular site visitors weighting to prioritize hyperlinks from articles with plenty of present natural site visitors.
It additionally auto-excludes any put up already linking to you, parsed from every candidate’s markdown physique, so that you’re not observing suggestions you’ve already used.
For every really useful article, the software additionally identifies the one paragraph most semantically aligned together with your new article. Then Claude Sonnet 4.6 drafts a pure 2-6 phrase anchor and rewrites that paragraph’s sentence to incorporate it, prepared to stick straight into the present article.
Starter immediate
Construct me an internal-linking software. Enter: both a broadcast weblog URL or pasted draft markdown for unpublished items. Embed the enter article with Gemini and cosine-compare in opposition to my pre-cached weblog put up vectors. Rescore prime candidates with authority weighting: 0.7 × similarity + 0.3 × log(org_traffic) — favours high-traffic hosts the place a hyperlink truly strikes rankings. Auto-exclude any host already linking to me (parse every candidate's markdown physique). For every prime host, establish the one paragraph most semantically aligned with the enter article — that is the place the hyperlink goes. Have Claude draft a pure 2-6 phrase anchor and rewrite a sentence within the host paragraph to incorporate it. Per-recommendation context: web page sim, passage sim, host's org_traffic / UR / refdomains, the host paragraph, and a one-line rationale. Cache passage vectors per host so repeat lookups are on the spot. Run lookups async with stay step standing; persist each lookup to historical past.
Ultimate ideas
If you happen to’re an Ahrefs buyer, you may strive Agent A without cost for one month.
Take a look at out a few of these prompts for inspiration, construct some functions and generate some studies, and see simply how a lot of the tedious elements of your job Agent A can sort out for you.
